(navigation image)
Home American Libraries | Canadian Libraries | Universal Library | Community Texts | Project Gutenberg | Children's Library | Biodiversity Heritage Library | Additional Collections
Search: Advanced Search
Anonymous User (login or join us)
Upload
See other formats

Full text of "Ultraviolet-assisted processing of dielectric thin films for metal oxide semiconductor applications"

ULTRAVIOLET-ASSISTED PROCESSING OF DIELECTRIC THIN FILMS FOR 
METAL OXIDE SEMICONDUCTOR APPLICATIONS 






By 
JOSHUA M. HOWARD 






A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL 

OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT 

OF THE REQUIREMENTS FOR THE DEGREE OF 

DOCTOR OF PHILOSOPHY 

UNIVERSITY OF FLORIDA 

■ 

2002 



Copyright 2002 

by 

Joshua M. Howard 









I would like to dedicate this work to the memory of my great grandmother, mormor, who 

will live on in my heart forever. 












ACKNOWLEDGMENTS 

I would like to first thank my advisor, Dr. Rajiv Singh, for all the enduring support 
and encouragement he has given to me since I joined his group. His guidance and 
friendship have played an integral role in molding me into the scientific researcher I am 
today. Next I would like to thank Dr. Valentin Cracuin, who also works with Dr. Singh 
as a visiting researcher from Romania. Valentin has acted as a great mentor throughout 
my time here. He has served as an inspiration for me to excel in all aspects of any project 
I undertake and taught me to look deeper than what may appear on the surface. Above 
all, Valentin has become someone whom I can truly call my friend. I would also like to 
thank Dr. Stephen Pearton for serving on my committee and for his keen insight into life 
that has always helped me put the world into perspective. I would also like to thank Dr. 
Cammy Abernathy for her smile and kind word that has often helped me through the day. 
Additionally, I thank Dr. David Norton for his assistance, enthusiasm, and for use of his 
laboratory equipment. Finally I would like to thank Dr. Fan Ren for use of his electrical 
characterization equipment and for serving on my committee. 

I would like to thank my parents, my brother David, and my extended family for all of 
the love and encouragement they have given me throughout my life. I will always 
cherish the active roles each played in my life growing up and how each person worked 
in a special way to help me become the person I am today. 

There are many contemporaries here at the University of Florida that deserve thanks 
for their help and friendship. In particular, I would like to thank Brent Gila, Mark 

iv 



Overberg, Michael Ollinger, Nabil Bassim, Francis Kelly, Anuranjan Srivastava, Srinivas 
Pietambaram, Chad Essary, and Seemant Rawal. I would also like to acknowledge all of 
the other wonderful group members current and past that are far too many to list. 

There are many friends that have come and gone throughout my life, but occasionally 
you meet people who you know will be there for the rest of your life. I would like to 
thank several of my dear high school friends, including Christopher Schmidt, Andrew 
Rogers, the whole Coombs clan, and Victoria Aronold, for all of their love, lessons, and 
friendship. I would also like to thank Daniel Kammler from my days at the University of 
Missouri-Rolla for being a truly amazing human being. 

Additionally, I would like to thank my best friend Jennifer Gray for all of her love and 
care. It is not every day that you get to be best friends with the person you love, but 
somehow I have found myself in this situation. Her confidence and spirit make my world 
a happier place to live. I thank her for everything that she is. 

It would be impossible to list all of the attributes that each of these close friends 
exhibits, but it can definitely be summed up that these people represent the best that the 
human race has to offer and for this I feel blessed that they have been a part of my life. 






TABLE OF CONTENTS 

page 

ACKNOWLEDGMENTS iv 

ABSTRACT xiv 

CHAPTER 

1 LITERATURE REVIEW 1 

Introduction 1 

The Transistor 1 

History 2 

Basic Transistor 2 

Fundamental Limitations 3 

Materials Selection Criteria 4 

Permittivity and Barrier Height 5 

Thermodynamic Stability 5 

Interface Quality 6 

Film Microstructure 7 

Gate Compatibility 7 

Process Compatibility 8 

Reliability 8 

Excimer Laser Basics 9 

Ultraviolet Lamp Basics 10 

Ultraviolet Excimer Radiation Sources 10 

Silent Barrier Discharge Excimer Radiation 1 1 

Materials and Properties 13 

Motivation 15 

Oxygenation of Films 18 

2 EXPERIMENTAL PROCEDURES 27 

System Geometries 27 

Laser System 27 

UVPLD System 29 

Excimer Annealing System 30 

Excimer Lamp Design 30 

Experimental Setups and Samples 31 

Interfacial Layer Formation 32 



VI 



In-Situ Ultraviolet PLD 32 

Post Deposition Ultraviolet Annealing 33 

Experimental Characterization Techniques 34 

Variable Angle Spectroscopic Ellipsometry 34 

X-ray Reflectivity 35 

X-ray Diffraction and Glancing Incidence X-ray Diffraction 36 

X-ray Photoelectron Spectroscopy 38 

Atomic Force Microscopy 38 

Fourier Transform Infrared Spectroscopy 39 

Electrical Characterization 39 

Current- Voltage Measurements 40 

Capacitance- Voltage Measurement 41 

Transmission Electron Microscopy 45 

3 INTERFACIAL LAYER FORMATION 59 

Anneal Conditions 59 

4 ULTRAVIOLET PROCESSING 81 

In-situ Ultraviolet PLD 81 

Barium Strontium Titanate (BST) 82 

Yttrium Oxide (Y 2 3 ) 90 

5 POST DEPOSITION ULTRAVIOLET ANNEALING 105 

Ultraviolet-Assisted Oxidation of Silicon 105 

H1O2 Post Deposition Anneal 106 

6 CONCLUSIONS 117 

APPENDLX ELECTRICAL DATA EXTRACTION METHOD 120 

LIST OF REFERENCES 125 

BIOGRAPHICAL SKETCH 129 












vn 



LIST OF TABLES 
Table page 



1-1 Industrial timeline for minimum feature size and equivalent dielectric thickness in 
MOSFET devices [2] 4 

1-2 Different wavelengths possible as a function of rare gas or rare gas halide mixture 

used for excimer decomposition 14 

1-3 Various properties of high-k oxide materials 15 

2-1 Conditions for growth and post deposition heat treatments of thin ZrC»2 films 32 

2-2 Conditions for in-situ ultraviolet annealing with an Hg lamp array during growth 

of BST and Y 2 3 thin films 33 

2-3 Conditions for post deposition excimer anneals of H1O2 thin films 34 

2-4 Conversion factors for series-parallel electrical equivalent circuits 44 

3-1 Partial pressure of oxygen for respective deposition ambients 60 

3-2 VASE thickness measurements of Z1O2 samples after post-deposition heat 

teatments in various ambients 61 

3-3 Thickness, roughness, and density data for the various XRR model options for the 

oxygen annealed Z1O2 62 

3-4 Modeling data for the as-deposited, vacuum, helium, and oxygen annealed 

samples 63 

4-1 XRR data for the PLD and UVPLD deposited BST samples 83 

4-2 XRD peak data for the PLD and UVPLD deposited BST samples 85 

4-3 GIXD peak data for the PLD and UVPLD deposited BST samples: 86 

4-4 XRR data for the PLD and UVPLD deposited Y 2 3 samples 91 

4-5 GIXD peak data for the PLD and UVPLD deposited Y 2 3 samples 92 



vin 






5-1 VASE thickness measurements for silicon oxidized with and without ultraviolet 

radiation at 3 00°C 106 

5-2 VASE thickness measurements for HfC>2 anneals with and without ultraviolet 

radiation 108 

5-3 XRR data for the thickness and density of the amorphous interfacial layer 109 



IX 



LIST OF FIGURES 
Figure page 



1-1 Schematic of a typical MOSFET. The gate, insulator, and silicon form the metal 

oxide semiconductor capacitor portion of a MOSFET 21 

1-2 Calculated conduction and valence band offsets for various perspective alternative 
dielectric materials on Si 22 

1 -3 Ternary phase diagrams illustrating a) "Si0 2 dominant," b) "no phase dominant," 
and c) "metal oxide dominant" systems. System c represents a stable condition 
for a metal oxide when in direct contact with silicon 23 

1-4 Schematic of total energy associated with laser ablation of a surface 24 

1-5 Sinusoidal voltage versus time plot indicating conditions where silent barrier 

microdischarges may occur 25 

1-6 Spectral emission characteristics for A) low pressure mercury lamp and B) xenon 

excimer lamp 26 

2-1 Ultraviolet-assisted pulsed laser deposition system, KrF excimer laser, and optic 

setup 46 

2-2 Homemade excimer annealing system equipped with vacuum ultraviolet lamp 47 

2-3 Schematic of excimer lamp illustrating the concentric tube design and how a radio 

frequency load is delivered to the system 48 

2-4 Tiny microdischarges from the ignited excimer lamp are the origin of the excimer 

radiation 49 

2-5 General x-ray reflectivity setup showing physical relationships between the 

acquired data and the modeling output 50 

2-6 General x-ray diffraction setup illustrating the interaction of x-rays with a 

structure as they pertain to Braggs Law 51 

2-7 Process of incoming radiation ejecting a characteristic photoelectron from a 

carbon sample 52 






x 



2-8 Schematic of atomic force microscope and the various components that allow up 

to atomic resolution 53 

2-9 Typical MOS capacitor prepared for this dissertation 54 

2-10 Block diagram of Keithley Win-82 system and how it connects to the probe 

station. Adapted from Keithley Win-82 operation manual 55 

2-1 1 Typical capacitance- voltage illustrating the three main regions that occur in a 
MOS device as a function of bias voltage applied. Adapted from the Keithley 
Win-82 operation manual 56 

2-12 A) parallel, B) series, and C) combined series and parallel models for generation 

of capacitance information 57 

2-13 Schematic of electron beam after passing through an ultrathin TEM sample. 

Some electrons are scattered while others remain unscattered 58 

3-1 XRR spectra of A) raw data, B) 1 layer model without fit, C) 1 layer model with 
fit, and D) 2 layer model with fit illustrating the importance of a good model 
when analyzing XRR data 71 

3-2 XRR spectra of A) as-deposited, B) vacuum annealed, C) helium annealed, and 

D) oxygen annealed Z1O2 thin films as modeled with the "2 layer model with fit"..72 

3-3 Plot of interfacial layer density as determined by XRR as a function of oxygen 

content in the annealing system. Bulk SiC»2 has been added as a reference 73 

3-4 Cross sectional TEM of a polycrystalline Z1O2 thin film atop an amorphous 

interfacial layer atop single crystalline silicon 74 

3-5 FTIR spectra of A) oxygen anneal, B) helium anneal, and C) as-deposited ZrC>2 
thin films showing the increase in Si — O bonding absorption in stretching, 
bending and rocking modes 75 

3-6 XPS data of Si 2p region of an as-deposited ZrC»2 thin film and after various post 

deposition anneals 76 

3-7 Plot of Si 2p binding energy of oxygen bonded to silicon as a function of oxygen 

content in the annealing system. Bulk SiC»2 has been added as a reference 77 

3-8 XPS spectra of A) raw data, B) 1 peak fit, C) 2 peak fit, and D) 3 peak fit fit 

illustrating the importance of a good model when analyzing XPS data 78 

3-9 XPS spectra with "3 peak fit" of A) as-deposited, B) vacuum annealed, C) helium 

annealed, and D) oxygen annealed Zr0 2 79 



XI 



3-10 Plot of O Is binding energy of oxygen bonded to silicon as a function of oxygen 

content in the annealing system. Bulk Si0 2 has been added as a reference 80 

4- 1 XRR spectra of A) UVPLD raw data, B) UVPLD 1 layer model, C) UVPLD 3 

layer model, and D) PLD 3 layer model for BST samples 94 

4-2 Cross sectional TEM of a polycrystalline BST thin film atop an amorphous 

interfacial layer atop a single crystal silicon substrate 95 

4-3 AFM of BST for PLD and UVPLD deposited samples. The UVPLD deposited 

sample exhibits increased roughness larger grain sizes 96 

4-4 GIXD of BST for PLD and UVPLD deposited samples. The UVPLD deposited 

sample exhibits increased (110) texturing 97 

4-5 Ols XPS of BST for PLD and UVPLD deposited samples. Peak B corresponds 
to the amount of physically trapped oxygen in the thin film structure and is 
reduced in the UVPLD deposited sample 98 

4-6 Current density versus voltage plot for BST of PLD and UVPLD deposited 

samples 99 

4-7 High frequency capacitance versus voltage plot for BST of PLD and UVPLD 

deposited samples 100 

4-8 XPS spectra of A) UVPLD raw data, B) UVPLD 3 layer model, C) PLD raw data, 

and D) PLD 3 layer model for Y 2 3 samples 101 

4-9 Cross sectional TEM of a polycrystalline Y 2 3 thin film atop an amorphous 

interfacial layer atop a single crystal silicon substrate 102 

4- 1 GIXD of Y 2 3 for PLD and UVPLD deposited samples. The UVPLD sample 

exhibits increased (222) texturing 103 

4-11 Ols XPS of Y 2 3 for PLD UVPLD deposited samples. Peak B corresponds to 

the amount of physically trapped oxygen within the thin film structure 104 

5-1 VASE thickness versus time results for oxidation of silicon with and without 

excimer radiation in oxygen 112 

5-2 VASE thickness of the overall Hf0 2 for samples annealed with and without 

utraviolet radiation 113 

5-3 XRR thickness measurements of the Hf0 2 interfacial layer for samples annealed 

with and without ultraviolet radiation 114 

5-4 XRR density measurements of the Hf0 2 interfacial layer for samples annealed 

with and without ultraviolet radiation 115 



Xll 



5-5 Capacitance- voltage measurements of H1O2 MOS devices for samples annealed 

with and without ultraviolet radiation 116 















xin 



Abstract of Dissertation Presented to the Graduate School 

of the University of Florida in Partial Fulfillment of the 

Requirements for the Degree of Doctor of Philosophy 

ULTRAVIOLET-ASSISTED PROCESSING OF 

DIELECTRIC THIN FILMS FOR METAL OXIDE 

SEMICONDUCTOR APPLICATIONS 

By 

Joshua M. Howard 

December, 2002 

Chair: Dr. Rajiv K. Singh 

Major Department: Materials Science and Engineering 

Oxygenation of oxide thin films was studied in this dissertation with application to 

metal oxide semiconductor technology. One of the main challenges envisioned for future 

microelectronic devices is finding a replacement gate dielectric for silicon dioxide (Si02) 

on silicon in metal oxide semiconductor applications. Several alternative high-k 

dielectric materials have been proposed as a solution to this problem because insertion of 

such a layer would allow thicker layers not subject to tunneling leakage issues 

encountered with ultrathin Si02 layers to be deposited. However, studies have indicated 

that new problems exist with these alternative layers. Among these issues are the 

formation of a low-k interfacial dielectric layer and an unacceptable level of oxygen 

vacancies in the films that contribute to leakage currents. In an attempt to address these 

issues, a three part experimental procedure has been designed to look at the role of 

oxygen in the thin film system. First, the formation of an interfacial layer as a function of 



xiv 






oxygen concentration in the thin film system has been investigated. Next, ultraviolet 
radiation was added to a pulsed laser deposition (PLD) system for in-situ ultraviolet- 
assisted PLD growth of dielectric films. Through the addition of ultraviolet radiation, 
highly reactive oxygen species are generated which alter the oxygenation conditions and 
dynamically alter the films properties and final oxygenation conditions. Finally, the 
application of ultraviolet annealing to already deposited hafnium dioxide (Hf0 2 ) thin 
films was studied to look at the possibility of more fully oxygenating the film after 
growth. A variety of characterization techniques indicate that both oxygen trapped in the 
film and oxygen that passes from the ambient through the film to the interface are 
responsible for the formation of the interfacial layer. Additionally, the application of 
ultraviolet radiation to a growth system can alter physical properties of a growing film, 
such as texturing of a polycrystalline film and overall reduction of trapped oxygen within 
the film. Furthermore, post-deposition ultraviolet annealing allows for the migration of 
oxygen species to the interface for growth of the interface and results in an unwanted 
reduction of the overall electrical properties. 









xv 



CHAPTER 1 
LITERATURE REVIEW 

Introduction 

Although silicon dioxide has remained the insulator of choice in metal oxide 
semiconductor devices for more than three decades, fundamental limitations of the 
material predict that its reign will soon come to an end. This has prompted many 
research teams around the world to search for a solution to this problem. However, the 
answer has proven to be surprisingly elusive. An approach chosen by many is to use a 
high dielectric constant material in place of the SiC>2 layer, thus alleviating the issues that 
are forcing the change in the first place. This dissertation addresses problems associated 
with the implementation of such an alternative layer. Special attention is given to 
exploring a second unwanted interfacial layer and developing the origins of that layer as a 
function of oxygen content in the system. Next, the effects of ultraviolet radiation are 
investigated to determine if a link could be made between the oxygen species in the 
system and the quality of in-situ and post-deposition processed samples. 

The Transistor 

The transistor is arguably one of the most important inventions of the twentieth 
century. It has paved the way for countless technological advancements as humans have 
moved into an era where communication around the world can occur in a matter of 
seconds. Data, information, and ideas can stream about the globe without time or 
distance ever coming into consideration. The transistor, as initially theorized and later 



developed, changed the course of computers forever. First it replaced vacuum tubes in 
early computers, and then was refined to a stage whereby fully integrated circuitry 
became a reality. Now it is integrated into microprocessors with millions of transistors 
working flawlessly on a single small microchip. The transistor truly has allowed humans 
to move into a new definition of existence. 
History 

In 1945 Bell Laboratories began research into a class of materials known as 
semiconductors. Their goal was to replace vacuum tubes (developed in 1906) for signal 
amplification in long distance telecommunications applications. The vacuum tubes were 
known to be unreliable, consume too much power, and produce too much heat. On 
December 16 th , 1947, the "point contact transistor" was created and consisted of strips of 
gold foil on a plastic triangle in intimate contact with a slab of germanium. This design 
was soon improved upon with the invention of the "junction (sandwich) transistor" a year 
later. This new transistor proved to be more rugged, practical and easier to manufacture. 
Twelve years later, a new design based on original theories for a field effect transistor 
was created. This design is the same one that has been continually used and adjusted for 
the past forty years. 
Basic Transistor 

A schematic of a modern metal oxide semiconductor field effect transistor (MOSFET) 
is shown in Figure 1-1 [1]. There are several important features to note. First, the 
transistor serves as a switching device that works on a field effect theory. As a voltage is 
applied to the gate, charge carriers from extrinsic dopants in the silicon are either 
attracted to or repelled from the silicon/silicon dioxide interface. This metal oxide 
semiconductor (MOS) layered structure is essentially a capacitor. Under certain 



conditions the excess charge carriers in the silicon may form a conductive channel at the 
interface. By sensing whether or not a current is flowing along this channel under 
certain given conditions, a one and zero interpretation may be generated. For example, if 
there is a current sensed, a one is interpreted and if no current is sensed, a zero is 
interpreted. These ones and zeros form the basis of all computer technology, which 
strings together long arrays of ones and zeros in binary coding. 

Fundamental Limitations 
To date, the microelectronics industry continues to increase performance by 
decreasing microelectronic device size. However, several roadblocks have been 
envisioned for the near future. The desire to increase metal oxide semiconductor field 
effect transistor (MOSFET) speed by decreasing the lateral dimensions (i.e., gate length) 
also demands a proportional reduction in the gate oxide thickness. The reduction in gate 
oxide thickness is necessary in order to maintain a constant capacitance per unit area in 
the semiconductor material. Within the next several years, silicon dioxide on silicon 
technology will reach fundamental limitations with respect to unacceptable leakage 
currents and low breakdown voltages. The problem occurs at very small gate thicknesses 
(sub 20A regime) required for future devices because of the exponential dependence of 
leakage current (due to direct electron tunneling) on the thickness of the dielectric layer 
[1-3]. Buchanan [2] illustrates the situation for Si0 2 , where at a gate bias of ~1 V, the 
leakage current changes from 1 X 10" 12 A/cm 2 at -35 A to 1 A/cm 2 at -15 A. This 
represents a change of twelve orders of magnitude in current for a thickness change of 
little more than a factor of two! This very high leakage current causes many concerns as 
to the operation of devices with respect to conceivably altering device performance, as 
well as creating standby power dissipation, reliability, and lifetime issues. As seen in 



Table 1-1, the industry timeline as presented by Buchanan [2] for the minimum feature 
size and the equivalent oxide thickness predicts quickly approaching deadlines for Si02 
technology. 



Table 1-1 Industrial timeline for minimum feature size and equivalent dielectric 
thickness in MOSFET devices [2]. 



Year 


Minumum Feature Size 
(tan) 


Equivalent Dielectric 
Thickness (A) 


1997 


0.25 


40-50 


1999 


0.18 


30-40 


2001 


0.15 


20-30 


2003 


0.13 


20-30 


2006 


0.10 


15-20 


2009 


0.07 


<15 


2012 


0.05 


<10 



Though the thinning process has been successful for the past four decades, limits are 
quickly being approached and must be addressed for the vigorous pace the 
microelectronic industry has followed for so many years to continue. Bear in mind that 
this must occur while still maintaining high quality standards that MOSFET technologies 
require. 

Materials Selection Criteria 

Due to the problem facing SiC>2 technology, there is a need for an alternate solution for 
SiC>2 as the dielectric of choice on silicon. However, as mentioned previously, to date, 
there are no immediate solutions to this problem. Though much work has been done, 
mainly in the area of high-k dielectric alternatives, incompatibilities with silicon 
(interfacial layer formation) as well as low quality deposited oxides with high leakage 
currents and large defect densities have continued to plague the insertion of an alternate 
layer. As outlined by Wilk et al. [4] in a thorough review of the current status on high-k 
dielectrics, there are a group of main considerations that must be satisfied before insertion 






of an alternate dielectric layer into silicon industrial processes will take place. In short, 
there are a set of main criteria that must be satisfied for a material to satisfactorily 
perform as a gate dielectric. 
Permittivity and Barrier Height 

First, the permittivity and barrier height must be taken into consideration. While 
many scientists originally felt that a material with a K > 25 would be necessary, it was 
later found that this may not be exactly true and that among other things, the barrier 
height of the material must also be taken into consideration along with the dielectric 
constant. An example of the band offsets for various high dielectric constant materials in 
contact with silicon is shown in Figure 1-2 [5]. The best materials for consideration 
should have a conduction band offset greater than ~1 eV, because as barrier height of the 
material decreases, there is also an exponential increase in leakage current as a result of 
Schottky emission of electrons into the conduction band [1,3]. Unfortunately, there 
exists a general relationship where barrier height decreases with increasing dielectric 
constant. A comparison of the band offsets for Si0 2 (e - 3.9) and BaTi0 3 (e > 2000) 
clearly illustrates this point. What this means is that there is an optimization problem to 
see which will have the larger effect. 
Thermodynamic Stability 

A second important consideration is the thermodynamic stability of the new dielectric 
material when in direct contact with silicon. An important approach toward predicting 
and understanding the relative stability of a particular three component system for device 
applications can be explained through ternary phase diagrams [6]. Hubbard describes the 
three main categories for phase diagrams as "Si0 2 dominant," "no phase dominant," and 
"metal oxide dominant." Each of these is shown in Figure 1-3 [6]. The metal oxide 



dominant type is of interest in this study. As an example, a comparison of the Zr-Si-0 
phase diagram (metal oxide dominant) versus the Ta-Si-0 (Si0 2 dominant) can reveal a 
great deal of information. Inspection of the Zr-Si-0 phase diagram in Figure 1-3 (C) 
shows tie lines that directly connect Zr0 2 and Si. This indicates stable ternary 
compounds (i.e., Zr0 2 should be stable on Si). Though sufficient data are unavailable for 
the Hf-Si-0 system, proximity on the periodic table and in similarities noted between Hf 
and Zr indicate that the same argument should hold true for the stability of Hf0 2 on Si. 
On the other hand, inspection of the Ta (or Ti) phase diagram in Figure 1-3 (A or B) does 
not have any tie lines connecting the compound of interest to Si, thus indicating a lack of 
stable ternary compounds. The stability of the systems, as predicted by these ties lines, is 
of paramount importance since it suggests that control of the dielectric/Si interface may 
be possible. However, it is important to keep in mind that even though the phase 
diagrams indicate stability in certain compounds over others, the phase diagrams are 
generated for equilibrium conditions. In fact, most of the deposition techniques do not 
occur under equilibrium conditions. This is an important difference to note since there 
have been many reports about systems that were predicted to have the thermodynamic 
considerations under control that still exhibit unwanted interfacial layers [7-10]. 
Interface Quality 

The third criterion that must be considered is the interface quality. This is of 
particular importance because an interface of poor quality will result in degradation of 
device performance. Imperfect bonding will cause under or overconstrained states that 
have a direct correlation to the leakage current and electron channel mobility and lead to 
degraded performance. In addition, a particularly rough interface can have adverse 
effects on device performance with respect to poor channel mobility characteristics. 



Film Microstructure 

An additional criterion which has been thoroughly analyzed is the film microstructure. 
There is still much debate on whether a polycrystalline, single crystalline, or amorphous 
film will work best. It has been argued that a polycrystalline setup is subject to high 
leakage currents due to nonstoichiometries that occur along the grain boundaries that act 
as leakage paths. This would seem to infer that a single crystal structure would be an 
ideal solution to this problem; however, the equipment typically used to grow single 
crystalline films is very expensive and not readily adaptable to a large scale industrial 
atmosphere. Furthermore, if it is possible to create a single crystal interface, it is very 
difficult to maintain the integrity of the single crystalline structure in subsequent thermal 
processing steps. One may also argue that the amorphous structure of SiC>2 that has been 
used so successfully for so many years would be ideal. This, however, also runs into 
problems since many of the high-k dielectric alternatives are not naturally in an 
amorphous state and quickly convert to a polycrystalline state when subsequent heat 
treatments of the layer are conducted. This particular concern has been an issue of great 
debate since many groups have generated decent data for each of the possibilities [11-13]. 
Gate Compatibility 

Next, gate compatibility must be taken into consideration. Namely, the dielectric has 
to be compatible with the current industry standard, which is a highly doped 
polycrystalline silicon (poly-Si) gate. This proves to be a much greater task than 
immediately apparent. There are issues associated with the diffusion of the dopant 
species into the gate dielectric materials, especially such materials as Zr0 2 and Hf0 2 that 
are reported to have rather open structures [2, 5]. The diffusion of the dopant ions into the 
dielectric may cause deleterious effects on the dielectric properties as well as create areas 



8 

in the gate where increased resistances are encountered. There is a possibility of solving 
this issue by replacing the poly-Si with a metal gate, but there are certain problems 
associated with that as well. First, the poly-Si gate currently allows tuning of the dopant 
levels to correlate to desired threshold voltages for both nMOS and pMOS devices. 
Switching to a metal means that this will no longer be an option, so either a midgap metal 
or two different metals with appropriate work functions will need to be implemented. It 
turns out that the midgap metal appears to be potentially limiting due to future predictions 
of the power requirements versus threshold voltages related to the midgap metal being to 
large and resulting difficulties turning the devices on. Midgap work function metal gate 
systems have also been predicted to be unable to provide a performance improvement 
worthy of the added process complexity [4]. This leaves the option of introducing two 
different metals into the CMOS system. Though this would require a bit of additional 
work to fully investigate the associated issues, this does appear to be a viable option for 
future devices. 
Process Compatibility 

Next, it is important to ensure that the dielectric does have good process compatibility. 
Simply stated, the microelectronics industry is tooled for large scale fabrication of Si0 2 
on silicon devices. It is easily conceivable that the dielectric that most easily slips into 
the current complex industrial process or requires the fewest number of modifications 
will be chosen. This is regardless of its status in other areas so long as continued device 
performance can still be achieved. 
Reliability 

Finally, the reliability of devices must be proven. Time dependent dielectric 
breakdown studies where higher voltages stress devices to breakdown conditions in 



reasonable time periods so that lifetime estimates for actual working conditions may be 
calculated must be conducted. Once done, the lifetime calculations must fit into the 
stringent ten year lifetime policy set up as the industry standard. To date, few reports 
have been generated of this nature and additional work must still be conducted. 

Excimer Laser Basics 

Though there are many techniques available for the deposition of thin oxide films onto 
a substrate, pulsed laser deposition has several key advantages that make it an ideal 
research tool. Stoichiometric transfer of molecules from the target to the substrate, 
effective rapid prototyping of many different materials, and a wide range of applications 
are foremost on this list. 

The ablation process itself is characterized by an input of energy from the laser to a 
given target material. Subsequently, the energy distribution shown in Figure 1-4 [14] 
occurs whereby the total energy for the system is represented as 

E = E r + E p +E d +E c (1-1) 

where E is the laser energy, E,- the reflected energy, E p the energy of the plasma plume, 
E d the energy of disintegration due to particles blown off by the vapor-gas jet, and E c the 
energy absorbed by the cavity wall. There are additional factors related to the laser 
energy that play an important role in surface response to pulse energy. For a given 
material, a combination of pulse energy and ablation threshold energy determines if 
ablation will occur, and to what level. It should be apparent that higher pulse energies 
result in an increasing level of ablation. If the pulse energy is less than the ablation 
threshold, material will not be physically removed from the surface (i.e., Ep and Ed tend 
to zero), but that energy may still be absorbed by the cavity wall (Ec) for laser annealing 



10 

experiments. All experimentation in this study was carried out at an energy value much 
greater than the ablation threshold. 

Ultraviolet Lamp Basics 
Ultraviolet Excimer Radiation Sources 

Over the past decade, numerous research teams have placed a considerable amount of 
effort and resources into the development of ultraviolet (UV) radiation sources that can 
be added to already existing processes related to the deposition and post deposition 
treatments of thin films. UV systems such as hollow cathode discharges [15, 16], 
constricted glow discharges [17], low pressure glow discharge from Hg lamps [18-22], 
and excimer lamps [15, 16, 23-35] have been used as additions to a variety of deposition 
techniques. Much research has been conducted with excimer lamps powered by radio 
frequency power supplies [23-29], but reports of microwave direct current magnetron 
power supplies have also been reported [36]. Radio frequency driven excimer based 
silent barrier discharge lamps have been the main contributors in a large part due to 
efforts by Boyd et al. Experiments include photo-induced deposition of silicon 
dielectrics such as silicon oxide, silicon nitride, and silicon oxynitride as well as direct 
oxidation of both Si and Ge substrates, and direct nitridation of Si surfaces [23, 24, 29]. 
High-k dielectric materials such as tantalum pentoxide (Ta 2 5 ) have been deposited via a 
photo-assisted CVD process [26, 27] whereby absorbtion of the highly energetic photons 
emitted by the UV source results in the direct photodissociation of the precursor 
materials. Along similar lines, post deposition excimer UV annealing of the as-deposited 
samples has been conducted, compared, and thoroughly analyzed [26, 27]. Additionally, 
the decomposition of palladium acetate films for subsequent electrodeless deposition of 
metals [37, 38], surface modification of polymeric materials [39], and photo-assisted 



11 

conversion of polyamic acid to form low-k dielectric polymeric polyimide [40] have also 
been explored. The majority of this work was conducted using excimer lamps. However 
several early experiments were conducted with a low pressure glow discharge Hg lamp 
[41] that has served as a useful tool for comparison in more recent excimer studies. 

Other groups have also investigated a variety of other UV radiation sources with 
respect to physical design and mechanisms for UV emission [15-17, 30-35]. Many of 
these are novel designs, but some are available from manufacturers (especially low 
pressure Hg lamps), and are discussed by the respective authors on a case by case basis. 
Deposition of silicon nitride on silicon by a low pressure Hg lamp [35], deposition of 
silicon nitride on III-V materials via a highly controlled photo-CVD process [19], and 
finally, the photolithic CVD of silicon dioxide utilizing an USHIO brand head on Xe 
excimer lamp [35] have also been pursued. Furthermore, Imai et al. [42] have looked 
into the densification of sol-gel films such as silicon dioxide and titanium dioxide by 
ultraviolet irradiation. 
Silent Barrier Discharge Excimer Radiation 

Since an in-house version of an excimer lamp was designed, built, and included in 
portions of the experimentation as an excimer UV radiation source, it is useful to look at 
a few of the properties associated with silent barrier discharges which seem to make this 
the best current option for a radiation source. First, with respect to other options 
available on the market, such as low pressure Hg lamps, the option of a Xe excimer lamp 
based on silent barrier discharge appears to be a superior option in several respects. The 
initial advantage lies in the power capabilities of the Xe lamp over the Hg lamp. 
Reported values for the Xe lamps range from as little as 10 mW/cm 2 to as great as 200 
mW/cm [24, 31, 35, 37, 43]. In comparison to low pressure Hg lamps, reported values 



12 

are less than 10 mW/cm 2 [28]. Intensities have been measured by a variety of methods, 
but commonly chemical actinometry or sodium salicylate scintillators coupled with 
monochromators have been used [28, 30, 36]. More recently, Ushio and Hamamatsu 
have made solid state photosensitive detectors for specific vacuum ultraviolet 
measurements. 

The reason that such high intensities may be achieved has to do with the physics 
behind the generation of a silent barrier discharge. Using xenon as an example, the gas 
undergoes an excitation as a response to a radio frequency signal. 

Xe+e-+Xe (1-2) 

Xe+Xe + Xe^>Xe 2 (1-3) 

Xe\ -^Xe + Xe + h v(l 12nm) ( 1 -4) 

The emission of radiation occurs in one of thousands of microdischarges that result as 
a function of radio frequency stimulus. As the sinusoidal oscillation of the voltage 
occurs, microdischarges may occur when a particular value for the voltage is achieved. 
This microdischarge remains until the change in voltage with respect to time is zero (i.e., 
dv/dt =0). This is shown graphically in Figure 1-5 [44]. 

Clearly, it is advantageous to have a higher flux of photons due to increased ability to 
dissociate various ambient gases (0 2 ), or precursor gases. Additionally, the greater 
photon flux may be advantageous with respect to the photonic effects, as the photons 
bombard the substrate and subsequently the growing film. The Xe excimer lamp has 
additional attractive features. First, the lamps are relatively inexpensive and easy to 
fabricate in comparison to other photo assisted processes where lasers or ion beams may 
be required [42]. The Xe lamp may also be easily adapted to very large areas by setting 



13 

up arrays of the lamps. This is particularly attractive for large scale applications in 
industry in the future if successful lab results support implementation. Additionally, the 
lamps are compatible with both atmospheric and vacuum operation. Xe excimers also 
exhibit narrow full with at half maximum (FWHM) values in single sharp emission 
spectra [23]. This is in contrast to the low pressure Hg lamps where a majority of the 
emitted radiation is actually occurring in a range above 200 nm which is above our range 
of interest. Both types of the emission spectra can be seen in Figure 1-6 [45]. To add to 
the low FWHM values, excimer lamps have a wide range of tunability with respect to the 
desired wavelength of radiation emitted. By simply changing the gas in the discharge 
region of the lamp, different wavelengths may easily be generated. This is illustrated in 
Table 1-2. 

As mentioned previously, the lamp design that Boyd, in conjunction with Kogelschatz, 
has chosen to use has been developed based off of United States patent 4,837,484 dating 
back to 1989 [46]. In this patent, a fairly detailed description of an excimer lamp design 
based off of silent barrier discharge is described and several variations are presented. In 
this work, reference is also made to "prior art" in the Soviet journal Zhurnal Prikladnoi 
Spektroskopii in a publication entitled, "Vacuum-ultraviolet Lamps with a Barrier 
Discharge in Inert Gases." From these, we have also made an in house version of the 
lamp. 

Materials and Properties 

As identified in the literature, there are numerous options for high-k dielectric 
materials. Possibilities include tantalum pentoxide (Ta 2 5 ), titanium oxide (Ti02), 
cerium oxide (Ce0 2 ), zirconium dioxide (Zr0 2 ), yttrium oxide (Y2O3), hafnium oxide 



14 



Table 1-2 Different wavelengths possible as a function of rare gas or rare 
mixture used for excimer decomposition. 


gas halide 


Excimer Complex 


Wavelength (nm) 


UV Range 


Ar 


126 


VUV 


Kr 2 * 


146 


F 2 * 


158 


ArBr* 


165 


Xe 2 * 


172 


ArCl* 


175 


KrI* 


190 


ArF* 


193 


KrBr* 


207 


UV-C 


KrCl* 


222 


KrF8 


248 


Xel* 


253 


Cl 2 * 


259 


XeBr* 


283 


UV-B 




Br* 


289 


XeCl* 


308 


I 2 


342 


UV-A 


XeF* 


351 



(Hf0 2 ), aluminum oxide (A1 2 3 ), and barium strontium titanate (Bao.sSro.sTiC^). As 
mentioned earlier, it is important to consider the main criteria associated with selecting a 
material. The main criteria that need to be taken into consideration for initial material 
consideration are permittivity, barrier height, and thermodynamic stability. As seen in 
Table 1-3, there are many options that are suitable for our investigation. 

Ta 2 5 , Ti0 2 , and BST all lack a thermodynamic stability in direct contact with Si. 
This can immediately eliminate them from consideration as MOS candidates. However, 
BST will be investigated due to its extremely large permittivity and previous experiments 
that suggest it may be a feasible material in light of certain findings to be explained later. 
From the list of other remaining materials, it is only a matter of selecting materials that 
are predicted to fulfill the other criteria. Input from industrial sources and extensive 
research conducted on these materials suggest that Y 2 3 , Zr0 2 , and Hf0 2 are excellent 



15 



Table 1-3 Various properties of high-k oxide materials 



M203 


9.3 


2.8 


3.97 


2054 


Basro.5Tio.503 


80 - 3600 


-0.1 -0.1 


6.02 


1625 


Ce0 2 


7 


— 


7.65 


2400 


H1D2 


-25 


1.5 


9.68 


2774 


Ta 2 5 


24-65 


0.3 


8.20 


1785 


Ti0 2 


80-170 


— 


4.23 


1843 


Y2O3 


10 


2.3 


5.03 


2439 


Zr0 2 


-25 


1.4 


5.68 


2710 



Oxide Dielectric Conduction Density Melting Stability in 

Material Constant Band Offset (g/cm 3 ) Point (°C) Contact with 

(eV) Silicon 

YES 

NO 

YES 

YES 

NO 

NO 

YES 

YES 

candidates. These, along with BST, have been used for various portions of the studies. 

Motivation 
Intense activity in the microelectronics field to solve this problem is currently 
underway. Several different approaches including new innovative device architecture 
designs (e.g., vertical structures and double gate planar transistors [4]), alternative gate 
oxides, and new process integration have all been proposed to solve this problem. To 
date, new device architectures are still under development and not ready to be 
implemented into full scale production. The possibility of moving the deposition of the 
final gate dielectric to the post heat treatment portion of the CMOS process is a daunting 
task with respect to companies adding to the already existing infrastructures. The option 
whereby the gate dielectric is replaced with a new dielectric material of higher 
permittivity has been pursued by many research teams [8, 9, 47-49]. According to the 
equations that govern metal oxide semiconductor (MOS) technology, if a dielectric 
material with a higher dielectric constant were substituted for the current Si0 2 oxide, a 
thicker layer could be synthesized while still maintaining a capacitance per unit area that 
is equivalent to a very thin Si0 2 layer. This effect can be seen in the equation 









high-k 



*■ high-k 
\ K ox J 



16 



U (1-5) 



where t is thickness and K is the dielectric constant of the respective high-k dielectric and 

the standard dielectric to be replaced. Another commonly referred to value is the 

equivalent oxide thickness (EOT) and signifies the thickness an equivalent Si0 2 layer 

would be if in fact the layer were Si02. This EOT value is paramount for research teams 

that desire to create a high-k gate stack that will be used in future MOS devices. This is 

because it is of no use to generate a layer that has an EOT thicker than if one were to just 

use Si02 in the first place. An additional equation that is easier to use to determine EOT 

when capacitance and gate area are known is 

___ 3.9xe n xA 

EOT = -^ (1-6) 

where 3.9 is the dielectric constant of a high quality Si0 2 , e is the permittivity of free 
space (8.854xl0" 12 F/m), A is the area of the gate, and C is the capacitance measured on 
the meter. This is simply a reorganized version of the more general equation for the 
capacitance of a MOS capacitor, where knowledge of any three of the four unknowns 
allows for calculation of the fourth 

n ss o A 

C=— (1-7) 

where all values are the same as above and e is the dielectric constant of the high-k 
dielectric. 

In theory, insertion of an alternative high-k dielectric layer is an attractive option, but 
in practice very difficult to achieve. The current Si0 2 dielectric layer for Si functions 
nearly ideally. Si is unreceptive to having that layer stripped and subsequent alternate 



17 

layers deposited in its place. There are problems with unwanted interfacial layers, which 
form during the deposition of the alternate dielectric layer, such as SiO x , or silicates. The 
formation of an interfacial layer has the ability to quickly nullify any of the beneficial 
effects of an alternative high-k layer. This is so because it changes the equivalent 
electrical circuit that would describe the MOS capacitor from a single capacitor system to 
a double capacitor system. The new system has a capacitance associated with the high 
dielectric constant material in series with a capacitance associated with the low dielectric 
constant material. This is then expressed as an overall capacitance by the following 
equation: 

1 1 1 



+■ 



(1-8) 



c c c 

w Total ** High-k •* Low-k 

Since the total capacitance is what would be the output on a measurement meter, a 
combination of this equation and equation 1-7 above, would allow for determination of 
the impact from each individual capacitance associated with both the high and low-k 
material. It should be easy to envision the desire to then apply these results a layer 
without any interfacial layer so that the total capacitance measured is 100% from the 
high-k layer with an end result of a much smaller EOT. 

If it is possible to eliminate the interfacial layer, there are still problems associated 
with unacceptable defect levels in the oxide and at the oxide/semiconductor interface. 
This combination of issues has prohibited the generation of high enough quality layers 
for viable substitution. 

In this research project, several of the more promising dielectric materials available 
have been selected and an attempt will be made to create a MOS capacitor with an 
alternate high-k material substituted in for the Si0 2 . The objective of this research is not 



18 

to generate an alternate high-k MOS stack with an extraordinarily low EOT, but rather 
study the role that oxygen has with respect to the interfacial region. That is, if a low 
dielectric constant interfacial layer forms, what is its chemical nature, and if oxygen took 
part in the formation, what were possible sources for the oxygen. Additionally, high 
intensity ultraviolet radiation sources will be added (during deposition and post 
deposition processes) to investigate the role of ultraviolet radiation as it applies to 
changes it causes in dry oxygen systems. The radiation used in in-situ deposition 
processes is supplied by an array of low pressure mercury lamps, while post deposition 
annealing radiation is supplied by a xenon excimer ultraviolet radiation source. It will be 
shown that deep ultraviolet radiation has a definitive role during each of the processing 
steps and that it can be an effective way to alter device properties such as structural, 
chemical and electrical characteristics. 

Oxygenation of Films 

Numerous reports have claimed a direct connection of the leakage current of alternate 
high-k dielectric materials to the amount of oxygen vacancies associated with the grown 
film [50-53]. It is also well known that while pulsed laser deposition is an excellent 
technique for the stoichiometric transfer of materials to the growing surface, like many 
deposition techniques, there is typically also an oxygen deficiency associated with the 
films. Traditionally, these films have undergone post deposition 2 treatments to help 
create more stoichiometric films and reduce the number of oxygen vacancies [54]. 

However, considerable effort has been placed on analyzing the effect of UV radiation 
as a method for creating more reactive oxygen species that will more effectively create 
stoichiometric structures and reduce oxygen vacancies. When radiation sources were 
added to conventional systems, photo-assisted growth of Si0 2 on silicon [20, 25, 29] 



19 

showed improvement with respect to enhanced oxidation rates and overall better 
properties, especially when the oxidations were conducted at temperatures much lower 
than those used in conventional dry oxidation. The addition of the low pressure Hg lamp 
to the system, which emits a majority of 254 nm radiation, but also a smaller percentage 
of 185 nm radiation [41], was claimed to have beneficial effects for the following reason. 
The radiation (especially the 185 nm portion) has the ability to convert normal dry 
oxygen (O2) in the reaction vessels into more reactive gaseous species. The species 
generated include ozone (O3) and other atomic oxygen species. This newly generated O3 
then undergoes a dissociation back to O2 and atomic O('D), ranging from several percent 
to greater than 10% in 2 [41]. Boyd et al. claim that the atomic O has the ability to 
move more easily through the growing Si0 2 matrix which enables it to reach the Si 
interface more readily for enhanced growth rates. It will also combine with defects that 
occur during normal growth of Si0 2 resulting in better stoichiometric higher quality 
films. 

According to later studies by Zhang et al, with photo-oxidation of silicon using a 
much higher intensity xenon excimer lamp, the 2 in the system follows the following 
scheme. The bond energy of 2 is known to be close to 5.1 eV and the energy, as 
calculated from the wavelength associated with Xe excimer emission, of the photons 
emitted from the xenon lamp are -7.2 eV. This allows the following reaction to take 
place. 

2 + hv(>.= 172nm)->0( 3 p) + 0( 1 D) (1-9) 

The oxygen atoms released can subsequently form ozone by the following reaction 

2 + 0( 3 p) + M -> 3 + M (M is a third body) (1-10) 



20 

where M is a third body participant and can be either a buffer gas, or in many cases, just 
the oxygen that is already in the system. The ozone can then be decomposed by further 
absorption of the vacuum ultraviolet light or thermally, thereby producing additional 
excited state *D oxygen atoms: 

3 + hv(X=172nm)^02 + 0('D) (1-11) 

The 0( D) atoms are claimed to be the main reason that enhancements are seen in 
samples that are processed with UV radiation due to their ease of moving about the 
matrices and/or their reactivity with defects inherent in the matrix. Additionally, the 
effect of the UV excimer radiation has also been adjusted as time progressed with respect 
to a second role of the UV radiation. This role is the effect of direct photonic 
bombardment of the surfaces from the energetic photons being emitted from the lamp. 
Wengenmair et al. [55] have shown in a study on TiN, where no oxygen is incorporated 
into the experiment, that UV radiation still caused differences in as-deposited samples 
versus UV deposited samples. This does serve as an indirect indication that photonic 
bombardment is playing a role in the UV process since no other process conditions were 
varied. Several different characterization techniques were used to confirm this 
observation. 









21 



OXIDE 






METAL 
ELECTRODE 




(SUBSTRATE BIAS) 



Figure 1-1 Schematic of a typical MOSFET [1]. The gate, insulator, and silicon form the 
metal oxide semiconductor capacitor portion of a MOSFET 



22 



6r 



9 

V 

c 

UJ 

-2 



■6 



3.5 



2.4 



3 . 08 

-0.1 



1.4 



1.5 



2.6 



1.3 



1.5 



1.1 



Si 



1.8 



4.4 



3.0 



2.3 



3.4 



3.3 



SJO, 



BaTIO, 

SUN Tl A "* 

S, 3 N 4 Ba2jO, 



3.4 



HfO, 



4.9 



3.6 



3.4 



Al,0 



Y 2 3 ZrSKD 4 






Figure 1 -2 Calculated conduction and valence band offsets for various perspective 
alternative dielectric materials on Si [5]. 









23 



M0 X /_\SiO a 




M 



JWSiy MSi, 
(C) 



Figure 1-3 Ternary phase diagrams illustrating a) "Si0 2 dominant," b) "no phase 
dominant," and c) "metal oxide dominant" systems. System c represents a stable 
condition for a metal oxide when in direct contact with silicon [6] 



24 



E=E r+ E p+ E d+ E c 



irradiated 
material 

3>- 




_ _ _. . ablation 
... ,. *depth 



- 




Figure 1-4 Schematic of total energy associated with laser ablation of a surface [15]. 






25 




Figure 1-5 Sinusoidal voltage versus time plot indicating conditions where silent barrier 
microdischarges may occur [44]. 












26 






Relative Spectral Energy Distribution 



100 



80 



2 60 

> 

J 40 



20 











(A) 
























p 








i £ 

2 S 


c c 
c c 
Q tn i 

— ■ — a j 




■ 546 nn 



■ L«L' JJ'J 300 400 500 bX 

Wavelength (Nanometers) 



•uj 



s: 



I 

• 60 

j- 

i 

i 

20 











(B) 






















£ 










c 
1 











100 jJ'J 300 400 500 :•:•: 

Wavelength (Nanometers) 



Figure 1-6 Spectral emission characteristics for A) low pressure mercury lamp and B) 
xenon excimer lamp. Plots adapted from [45] 



CHAPTER 2 
EXPERIMENTAL PROCEDURES 

System Geometries 

There are four main components used for experimentation. This includes the laser 
used for the laser ablation process, a traditional pulsed laser deposition (PLD) system that 
has been modified to an ultraviolet-assisted PLD (UVPLD) system and a homemade 
vacuum system equipped with a silent barrier discharge excimer lamp that is used for 
post deposition annealing of grown high-k dielectrics. The excimer lamp used for the 
post deposition system will be discussed in detail. 
Laser System 

A Lambda Physik LPX 305 i KrF excimer laser (s/n 9412 E 4188) was used for all 
laser ablation portions of the experimentation. This particular laser works in a pulsed 
mode delivering 25 nanosecond duration square wave shaped pulses at frequencies 
ranging from 1 - 50 Hz and output energies from 10 - 1 100 mj (fluence 0-3 J/cm 2 ). The 
computer is triggered by either an "internal" computer or from a remote "external" 
computer. 

Excimer lasers work on the principle of stimulated emission of photons in a cavity 
where Kr, F, and a buffer gas are all contained. The Kr and F are elevated into excited 
states by application of very high voltages (16 -21 kV) so that excited KrF* complexes 
form and upon their decay give off single wavelength 248 nm radiation. As a result of 
specific cavity design, conditions exist whereby stimulated emission of coherent radiation 



27 



28 

occurs and subsequent amplified high energy laser output in a pulsed mode can be 
obtained. 

The laser radiation emitted from the cavity is directed through a series of lenses and an 
aperture until it finally impinges the sample. For the PLD system, the laser beam first 
traverses a collimation lens in an attempt to keep the radiation from diverging due to 
scattering as it passes through the ambient air. Immediately after the collimating lens, a 1 
x 2 cm aperture is used to cut away the more diffuse lower energy edges from the 
incident beam. The beam then progresses into the chamber through an excimer grade 
fused silica window and then through a focusing optic with a focal length of 25 cm. 
Finally, a 2 mm thick quartz plate is used to protect the optic from material ablated from 
the target. Using a Gentec Sun Series EMI energy meter (s/n 86052), the energy inside 
the chamber was determined after the pulsed laser beam had passed through all the 
optical components (without the focusing lens). An additional 10% of that was 
subtracted to take into account for the final focusing optic. From this value, divided by 
the area of the final spot size on the target, a precise calculation of the laser fluence was 
possible. 

The ablation spot on the target had dimensions of 2 x 5 mm in a nearly perfect 
representation of the rectangular aperture used earlier in the beam path. This type of spot 
was achieved by adjusting the focusing lens position to a local greater than the focal point 
of 25 cm to a position -35 cm, which coincided with the image plane of the lens. 
Because the aperture was used, imaging of it was possible and a spot with a highly 
uniform energy density was created. It was found that while ablation at the focal point 
resulted in a smaller spot size corresponding to a greater fluence, this was unnecessary 



29 

and detrimental to the setup. First, it was unnecessary because the threshold value for 
laser ablation of oxide materials frequently does not require high energy densities, as was 
the case in our experimentation. Second, taking the beam to the focal point resulted in a 
highly irregular spot geometry which also consisted of two additional satellite peaks to 
either side of the main ablation spot. The main spot was characterized by a Gaussian 
type energy distribution while the satellite peaks were clearly of a different energy all 
together. All samples grown in the PLD setup used this laser and optical setup. 
UVPLD System 

The pulsed laser deposition performed for this dissertation was done in a Neocera 
brand vacuum system. The entire system with laser, but without computer control is 
shown schematically in Figure 2-1. The system is a single chamber design that is 
routinely backfilled with nitrogen to atmospheric pressure so samples may be mounted 
and/or removed. It can easily reach vacuums of 1X1 0" 6 Torr within an hour, IX 10" 7 Torr 
within twenty four hours, and 1X1 0" 8 Torr with the addition of liquid nitrogen cooling to 
the system. Vacuum is achieved via a Pfeiffer MD-4T oil free diaphragm roughing pump 
and a Pfeiffer TMU 230 turbo pump. A calibrated Neocera brand stainless steel resistive 
heater capable of 850°C is mounted vertically in the chamber and used to controllably 
heat and cool the substrate to and from the desired temperature. There is a computer 
controlled multitarget carousel available for depositions of up to six different materials 
for multilayer or superlattice experiments. An array of low pressure Hg lamps has been 
added to the conventional PLD setup to convert it into a UVPLD apparatus. Ultrapure 
gases may be added to the system through a highly sensitive Varian brand leak valve for 
a wide range of deposition ambients and pressures. 






30 

Excimer Annealing System 

An in house vacuum system, seen in Figure 2-2, was setup for the purpose of post 
deposition annealing of samples. The system has the ability to reach high vacuum 
(1X10" Torr) conditions via a Varian SD-450 two stage rotary vane roughing pump 
using Fomblin vacuum pump fluid coupled with a Pfeiffer TPU 1 70 turbo pump. There 
is precise atmospheric control via a highly sensitive Varian brand leak valve, and it is 
equipped with an Excel Instruments stainless steel resistive heater capable of up to 850 
°C. The main feature of the system is an excimer lamp designed and built in-house that 
has been added to the system. The lamp was setup in a through mode so that the 
cylindrical quartz tubing of the lamp entered the vacuum chamber on one side through 
vacuum feedthroughs, traversed the entire vacuum cavity and exited another vacuum 
feedthrough out the other side. This is important in the design because it allows for 
efficient water cooling of the lamp as well as vacuum compatibility. The system is 
designed to accommodate full evacuation of essential portions of the excimer lamp (1x10" 

Torr) and then, through a gas manifold, precisely backfill the lamp to any desired 
pressure up to atmospheric pressure with any gas of choice. 
Excimer Lamp Design 

The lamp design is based off of United States patent 4,837,484 (1989) and similar 
designs employed by Boyd et al. in London, England. The basic design of the lamp, as 
seen in Figure 2-3, entails the fusion of a pair of quartz tubes. The tubes are made from 
ultrahigh purity quartz (Suprasil) produced by Heraeus, and is the only quartz in the 
world capable of passing -80% of the excimer ultraviolet radiation emitted by xenon 
complexes (172 nm). Standard high quality quartz will not pass any of the radiation at 
this wavelength. The quartz tubes are of two diameters so that one longer tube can fit 



31 

concentrically within the other, thus forming a gap. The ends of the shorter tube are 
fused to the longer tube in a way such that a constant gap along the length is maintained. 
An additional third tube (the gas inlet in Figure 2-3) must also be added to the tubes so 
that there is an access port to the gap. This tube will later be used as a port for pulling a 
high vacuum on the gap, and eventually for backfilling of any of a variety of gases 
capable of excimer decomposition. Table 1-2 shows a list of possible gases that may be 
used in a given lamp and the radiation wavelength they emit. The outer tube is then 
covered by a metallic mesh. The finer the mesh the better since there will be less 
obstruction for exiting radiation and therefore result in greater energy densities. This 
mesh is to be grounded. When mounted, the inside of the inner tube will serve two 
purposes. First, deionized water will flow through the system at all times during 
operation as an efficient cooling mechanism. As the power input into the system 
increases, the cooling water plays an increasing important role. Second, a metal wire 
capable of carrying high powers will be spiraled along the interior of the tube. This wire 
is connected to a radio frequency (RF) power supply (T & C Conversions, Inc.), that 
generates up to a 200 watt load to induce excimer decomposition. Figure 2-4 shows the 
microdischarges of the ignited excimer lamp located in the center vacuum portion of the 
high vacuum system. 

Experimental Setups and Samples 

There are three main areas that comprise the work discussed in this dissertation. As a 
result, it is important to delineate the various samples that were produced or modified in 
each stage, and discuss any preparations or modifications that were completed. 



32 

Interfacial Layer Formation 

In this initial portion of the experimentation, the characteristics of an interfacial layer 
were investigated. Ultrathin Zr0 2 samples were deposited in a conventional PLD 
chamber and exposed to different ambient temperatures in a post deposition anneal. 
Vacuum, helium, and oxygen were used as annealing ambient gases. The main goal for 
this investigation was to try to determine the source or sources of oxygen that may exist 
during a typical deposition and, in particular, how they impact the characteristics of an 
unwanted interfacial layer. The conditions for deposition and post deposition heat 
treatments are shown in Table 2-1. The samples were investigated by variable angle 
spectroscopic ellipsometry, x-ray reflectivity, x-ray photoelectron spectroscopy, and 
Fourier transform infrared spectroscopy. These techniques will be discussed later in this 
chapter. 

Table 2-1 Conditions for growth and post deposition heat treatments of thin ZrQ 2 films 



Sample 



Post Deposition Anneal Atmosphere 



No Anneal 



Vacuum (6 x 10' 6 Torr) 



Helium (600 Torr) 



Oxygen (600 Torr) 



In-Situ Ultraviolet PLD 

For this experimentation, a conventional PLD system has been fitted with an array of 
Hg lamps. The lamps emitted radiation in the ultraviolet and deep ultraviolet regions. 
That is, a majority of the radiation was emitted at -256 nm, and a much smaller portion 
(-10%) at -185 nm. The 185 nm radiation is responsible for the conversion of oxygen 
into ozone and other atomic oxygen species. Once the lamps were added to the system, 
Y 2 3 and Bao.5Sr . 5 Ti03 (BST) thin films were deposited. An optimum pressure of 



33 

ambient oxygen during ablation was determined for each of the materials. The 
optimization was done by comparing the full width at half maximum (FWHM) of a given 
x-ray diffraction peak of films deposited at different pressures. Table 2-2 shows the 
specific oxygen conditions used for the depositions. 



Table 2-2 Conditions for in-situ ultraviolet annealing with an Hg lamp array during 
growth of BST and Y2O3 thin films. 





Oxygen Pressure (mTorr) 


Ultraviolet Radiation 


Barium Strontium Titanate 






1 


20 


NO 


2 


10 


YES 


Yttrium Oxide 






1 


10 


NO 


2 


4 


YES 



Once samples were obtained, they were characterized by variable angle spectroscopic 
ellipsometry, x-ray reflectivity, x-ray diffraction (glancing incidence x-ray diffraction), 
atomic force microscopy, transmission electron microscopy, current-voltage, capacitance 
voltage, and x-ray photoelectron spectroscopy. 
Post Deposition Ultraviolet Annealing 

In the final portion of experimentation, an array of hafnium dioxide (H1O2) samples 
underwent a post-deposition excimer annealing step in the system shown in Figure 2-2. 
These films were deposited by a chemical vapor deposition technique by the Motorola 
Company on a 200mm p-type silicon wafer. The wafer was broken into smaller pieces 
and then placed in the excimer annealing chamber. The main difference with the excimer 
radiation compared to Hg ultraviolet radiation is that the emission in the ultraviolet 
region is a focused single sharp emission peak at 172 nm. Therefore, all of the radiation 
is acting to convert the dry oxygen into ozone and atomic oxygen. Table 2-3 shows the 
conditions for the post deposition anneals. 



34 



Table 2-3 


Conditions for post deposition excimer anneals o 


f Hf0 2 thin films 




Anneal Temperature °C 


Excimer Ultraviolet 
Radiation 


1 


No Anneal 




2 


300 


NO 


3 


300 


YES 


4 


475 


NO 


5 


475 


YES 


6 


600 


NO 


7 


600 


YES 



One set of samples was annealed in dry oxygen at temperatures of 300, 475, and 600°C. 
A second set of samples was grown in the presence of dry oxygen and excimer UV 
radiation, again at 300, 475, and 600°C. Both systems were at an oxygen pressure of 5 
mTorr. An as-deposited control sample from the Motorola wafer that did not undergo 
any post deposition annealing was also analyzed. Upon annealing, the samples were 
tested by variable angel spectroscopic ellipsometry, x-ray reflectivity, x-ray diffraction 
(glancing incidence x-ray diffraction), atomic force microscopy, and capacitance voltage 
measurements. 

Experimental Characterization Techniques 

Several measurement techniques have been used to illuminate the differences seen 
between the samples that underwent radiation treatments versus those that did not. The 
properties of the films that were examined included structural, chemical, and electrical 
properties. Additional information is provided for the electrical characterization in the 
appendix due to its importance with respect to the ultimate goal of creating a better 
transistor. 
Variable Angle Spectroscopic Ellipsometry 

VASE is a very powerful, simple, and nondestructive means for determination of 
thickness in multilayered structures as well as optical properties such as index of 



35 

refraction and extinction coefficient. The optimum thickness range for ellipsometry is 
between 1-1000 nm and is well suited to the flat planar materials with low surface 
roughnesses generated for this dissertation. All measurements were made with a J. A. 
Woollam brand M-88 variable angle ellipsometer, with the angle of incidence set to 75°. 
This angle is optimum for semiconductor and microelectronic thin films. In this 
technique, the sample is subjected to a collimated beam of light from a xenon lamp. The 
light is adjusted so there is a known polarization state as is leaves the polarizer. After the 
beam of know polarization interacts with the sample, it will then exhibit a new 
polarization state. This new polarization state is interpreted by an analyzer, allowing a 
determination of the ratio of the complex Fresnel reflection coefficients and ultimately 
the psi and delta functions that are related to the properties of the sample. This method 
then requires the experimentally determined data to be compared to data from a 
theoretical model that is adjusted until a "best fit" can be found between the two. This 
does indicate that the better the theoretical model used for fitting, the closer the answer 
will be to the correct values. Since there is also the possibility that several different 
theoretical models fit well to one set of experimental data and may produce equivalently 
good fits, it is up to the user to evaluate and assimilate all information about the sample to 
develop a physically realistic theoretical model. It is therefore useful to make 
measurements at optimum wavelength and angle of incidence combinations to keep the 
assumed theoretical model simple yet realistic. 
X-ray Reflectivity 

X-ray reflectivity (XRR) measurements were made with the assistance of a Philips 
brand MRD X'Pert system. Data generated from an x-ray reflectivity plot include 
thickness, roughness, and density of a given material. Figure 2-5 is an example of a 



36 



typical plot showing how the different features of an acquired spectrum relate to the 
various data that can be extracted from the spectrum. Similar to VASE, the technique is 
well suited to smooth flat samples with low surface roughnesses and is most sensitive to 
samples in the 2-400 nm thickness range. This particular x-ray based analysis technique 
works by impinging the sample with x-rays over an array of angles ranging from slightly 
sub-critical angles to the first few degrees after the critical angle. The critical angle, also 
known as Brewster's angle, corresponds to the point at which x-rays change from total 
reflection off the surface to absorption and interaction with the sample as defined by 
Snell's law. The retrieved data results from monitoring the intensity of the x-ray beam 
reflected from various interfaces relative to the incident beam as a function of the 
scattering transfer vector. Fresnel equations will then describe the interaction of the x- 
rays with one another and with interfaces encountered in the structure. The constructive 
or destructive nature of the x-rays at a given angle results in generation of a fringe 
pattern. Similar to VASE, this pattern is compared to a user input model that 
incorporates thickness, roughness, density, and absorption parameters. Again, the better 
the input model, the more accurately the results will match the real physical structure. 
X-ray Diffraction and Glancing Incidence X-ray Diffraction 

XRD is most commonly used to identify crystalline phases and measure the structural 
properties of the phases such as strain, grain size, epitaxial quality, phase composition, 
preferred orientation, and defect structure. Additionally, the technique is noncontact and 
nondestructive and will produce spectra for films as little as 50A in thickness. The basic 
setup of the diffraction process is seen in Figure 2-6. The premise of this relatively 
simple characterization tool is the generation of diffraction peaks due to constructive and 
destructive interference from x-rays scattered by the atomic planes in a crystal. The 



37 

condition for constructive interference from planes with a given spacing is given by 
Bragg 's Law: 

\=2d h kisin0 hk i (2-1) 

where Xis the wavelength of the incident x-ray radiation (typically Cu Ka), d h ki is the 
d-spacing between (hkl) planes, and h ki is the angle between the atomic plane and the 
incidence direction for the x-rays. For single crystal films, there is only one specimen 
orientation that will satisfy the conditions for Bragg diffraction, however, with thin films 
that are polycrystalline, fiber textured, or exhibit preferred orientations, several families 
of planes may contribute to a diffraction system. X-ray diffraction data for this 
dissertation was obtained with a Philips MRD diffraction system. Once an x-ray 
diffraction pattern is generated, positive phase identification can be achieved by 
comparing measured d-spacing from the diffraction pattern (and their integrated 
intensities) to a known JCPDS powder diffraction standard. 

In certain cases, it is not possible to get any type of diffraction pattern from a sample 
because the sample is either too thin or the peaks of interest are being masked by a much 
more intense peak that comes from the single crystal substrate. In this instance, a special 
mode of x-ray diffraction known as glancing incidence x-ray diffraction may be used. In 
this setup, the incident angle of the x-ray system is fixed at a small value (typically from 
0.250-1.000°) and the receiving slit is allowed to scan through a typical range for a 
conventional XRD 2d scan. As a result, only planes that satisfy the Bragg condition with 
the additional constraint in place will produce peaks. In a polycrystalline sample with 
many orientations, this will still generate a representative x-ray diffraction plot, but with 
much higher surface sensitivity and under proper conditions, without any masking peaks 



38 

from the silicon substrate. This method proved to be beneficial for several samples in 

this dissertation due to the low thicknesses used in different portions of the 

experimentation. 

X-ray Photoelectron Spectroscopy 

XPS spectra were collected using a Perkin Elmer 5100 installation using Mg Ka 
radiation at a takeoff angle of 90°. This technique is clearly one of the most broadly 
applicable general surface analysis techniques due to its high surface sensitivity and 
quantitative chemical state analysis capabilities. Elemental detection includes all 
elements except hydrogen and helium. Again, the smooth flat samples measured in this 
dissertation are optimum for this analysis technique. In the XPS process, high energy 
photons can ionize atoms to produce free electrons known as photoelectrons. The kinetic 
energy (KE) of the electron depends on the energy of the photon by Einstein's 
photoelectric law: 

KE = hv - BE (2-2) 

where hu is the energy of the incident photon and BE is the binding energy. In this 
equation, hv is known, KE is measured, and BE is therefore the determined output. The 
measured BE is specific to the atom concerned and thereby succinctly identifies the atom. 
Figure 2-7 shows the photoelectron process with carbon as an example. While the Mg 
Kahas sufficient energy (-1486.6 eV) to eject the innermost electrons from carbon, the 
photons may also remove the 2s or 2p electrons. XPS data may be interpreted by the user 
to give sensitive elemental and chemical state analysis. 
Atomic Force Microscopy 

Atomic force microscopy measurements were made with a Digital Instruments brand 
Nanoscope III operating in tapping mode. This technique was chosen for its exceptional 






39 



ability to produce topographic images of a surface in all three dimensions at remarkably 
high resolutions. If conditions are properly set, atomic resolution is attainable. This 
technique is also perfectly suited to the insulating, low roughness samples created for this 
dissertation. In an atomic force microscope, a sharp tip is mounted on a flexible 
cantilever. When the tip comes into close proximity of a sample surface, van der Waal 
forces repel the tip causing the cantilever to deflect. A piezoelectric scanner is 
responsible for moving the cantilever/tip assembly along the surface of the sample, while 
a laser reflecting off the end of the cantilever maps the topographical changes the 
cantilever senses. This is shown schematically in Figure 2-8. 
Fourier Transform Infrared Spectroscopy 

This method is one of the few techniques that provides information about the chemical 
bonding in a material, and is nondestructive. Here the goal is to determine changes in the 
intensity of a beam of infrared radiation as a function as a function of wavelength or 
frequency after it interacts with a sample. The FTIR used in these investigations is a 
Nicloet MAGNA 760 equipped with potassium bromide (KBr) optics and operating in 
transmission mode. The main feature is the ability to determine in a qualitative (and 
quantitative) sense the types of bonds that are present in a thin film structure. 

Electrical Characterization 

A primary tool for indication of quality of thin film is the use of electrical 
characterization techniques. After deposition or processing of thin films, metal oxide 
semiconductor (MOS) devices were fabricated and measured. Typical preparation of the 
film included the deposition of either gold (Au) or platinum (Pt) contacts via evaporation 
or DC magnetron sputtering, respectively. In both cases, a mask with an array of circular 
dots ranging from 25-500um was used to create dot arrays on the samples. The backside 



40 

contact was either evaporated Au or silver (Ag). In both cases, the backside of the wafer 
was cleaned and the surface was abraded so that clean silicon, without any native silicon 
dioxide, was present. It is of paramount importance to understand the consequences of 
which type of metal is chosen depending on what dielectric is deposited and whether the 
silicon substrate is n-type or p-type. If improper metals are chosen for a given setup, 
band alignment conditions may exist whereby non-ohmic contacts are created within the 
device itself. As an example, the backside metal contact on a p-type silicon wafer should 
have a work function greater than that of the silicon. If these conditions are met, there 
should be an ohmic contact, if not, the contact will be a rectifying Schottky contact. The 
most recent generation of a MOS capacitor fabricated for this dissertation is shown in 
Figure 2-9. After metallization of the front side contact (i.e., gate) was complete, 
samples underwent heat treatments either in an AG Associates rapid thermal annealing 
(RTA) furnace or in a conventional furnace. In both cases, in a forming gas atmosphere 
(10% H 2 , balance N 2 ) was utilized. Both current-voltage and capacitance- voltage 
measurements were taken once the MOS capacitors were formed. 

Current- Voltage Measurements 

Once MOS devices were fabricated, a Keithley Instruments Inc., KI236 source 
measurement unit (SMU) was used to measure the current flow through a device. The 
236 SMU was attached to a black box probe station equipped with a pair of Signatone 
Inc, micromanipulators. The micromanipulators were fitted with tungsten probes that 
were milled to produce a fine ~5 um tip. The output of the SMU was connected to the 
micromanipulator in contact with the gate (Pt or Au) while the input was connected to the 
backside contact (Ag or Au). This configuration is optimal for the determination of 
current leakage pathways directly below the device being measured (i.e., it avoids stray 



41 

leakage paths). Determination of the leakage current is an essential first step in analyzing 
an MOS device. Typically a current compliance threshold (a value that may not be 
exceeded) of 100 nA is input into the measurement parameters. Then a direct current 
bias sweep is conducted over a voltage range (usually negative to positive) and the 
amount of current that passes through the MOS structure is monitored. Typically it is 
best to start at a small voltage sweep range so that it is possible to determine if the device 
is leaky without causing a large amount of bias induced defects. The main goal of the 
leakage current measurement is to identify a high quality device that can be used for 
capacitance voltage measurements and to determine if excessive leakage is present that 
needs further investigation. 
Capacitance-Voltage Measurement 

Capacitance-voltage measurements serve as one of the most versatile and sensitive of 
all electrical characterization techniques. It is the ultimate tool for determining discreet 
differences in a MOS device that may serve as the final word in whether a given 
processing condition has resulted in a high enough quality device to apply to MOSFET 
applications. These measurements were carried out with a Kiethley Instruments Inc. 
Win-82 measurement system. This system, as seen in Figure 2-10 is comprised of four 
main components that work in unison. The Keithly 590 capacitance meter is used for 
high frequency capacitance measurements at 100 kHz and 1 MHz. The Keithley 595 
capacitor is used for simultaneous quasistatic low frequency measurements. The 
Keithley 230 voltage source is used for static bias condition measurements. These three 
devices are wired into the Keithley 5951 remote input coupler, which serves to filter the 
device data to and from the various pieces of equipment. Due to its ability to give 



42 

detailed information about the quality of the MOS device, the Win-82 system will be 
discussed in the appendix. 

The output of a typical CV curve can be seen in Figure 2-11. There are several 
important features that should be noted. First, there are three important regions with 
respect to gate bias voltage to take into consideration, known as accumulation, depletion, 
and inversion. The presence of the different regions is a result of the majority charge 
carriers (e.g., holes in a p-type silicon wafer) in the semiconductor. When a negative bias 
is applied to the gate electrode, positively charged holes are attracted from the 
semiconductor bulk region to the oxide/semiconductor interface where they accumulate 
(the accumulation region). The depletion region is generated when a gate is made less 
negative and the reduced field across the oxide allows the charge at the interface to 
diminish. As the sign on the voltage changes from a negative to a positive, majority 
carriers are repelled from the interface creating an area depleted of majority carriers (the 
depletion region). Finally, the inversion region is generated when the voltage becomes 
very positive and the depletion width has increased to a point where other mechanisms 
may become important. For example, in the depletion region, the product of the 
concentration of electrons and hole (np) is much less than the square of the intrinsic 
carrier concentration (n ( 2 ) and in this case, pair generation may occur and the subsequent 
minority carriers may migrate to the interface. The result of this is prevention of further 
depletion and a constant value for the capacitance. An additional possible scenario may 
occur whereby insufficient time is allowed for pair generation in which case a model 
called deep depletion will occur. 



43 

The high frequency measurement system has the ability to measure two different 
frequencies and using Metrics ICS software, two different equivalent circuit models, seen 
in Figure 2-12, to generate capacitance values. There are parallel and series models that 
represent two different physical structures. The series model addresses a capacitor in 
series with a resistance of some sort (possibly from the semiconductor). The equation 
that describes this scenario is 

Z = R + iX (2-3) 

where Z is the impedance, R is the resistance, and X is the reactance. Additionally, the 
parallel model addresses a capacitor in parallel with some sort of conductance (possibly 
leakage through an ultrathin film). The equation that describes this scenario is 

Y = G + iB (2-4) 

where Y is the admittance, G is the conductance, and B is susceptance. Note that the 
admittance, conductance, and susceptance are each the reciprocals of the impedance, 
resistance, and reactance, respectively. The reactance and susceptance can then be 
further described in a capacitive sense as: 

X — (2-5) 



coC s 






Y = wC P (2-6) 

where o> is the frequency of the setup and C s and C P are the capacitance when using the 
series model and parallel model respectively. The net impedance of the equivalent series 
and parallel circuits at a given frequency are equal, but the individual components are 
not: 

K + iX = — — (2-7) 

G + iB 



44 

If there is a lossless circuit (i.e. R = and G = 0) the C s and O are equal. However, 
since circuits do have losses, a dissipation factor is added to the system. 



D = aC s R 



D = 



Q)C, 



(2-8) 



(2-9) 



Through further numerical manipulation, it is determined that conversion from one 
equivalent circuit model to the other is readily possible as seen in Table 2-4 



Table 2-4 Conversion factors for series-para 



Model 



Parallel C P , G 



Series Cs, R 



Dissipation Factor 



Q coC f 



D = — = coC<.R 
Q 



lei electrical equivalent circuits 



Capacitance 



C s =(\ + D 2 )C f 






l + D' 



Resistance or 
Conductance 



R = 



D 2 



(\ + D 2 )G 



G = 



D< 



(\ + D 2 )R 



This treatment has been applied to illustrate how capacitance values are generated 
from the impedance values and subsequent conversion to series and parallel cases. As 
mentioned earlier, the series represents a physical scenario where a capacitor is in line 
with some type of resistance, while the parallel mode represents a physical scenario 
where a capacitor is in parallel with some type of conductance. This however does not 
take into consideration the physical possibility of an ultrathin film that is leaky, but also 
encounters resistance from the substrate. In this case, a three element electrical circuit, 
also seen in Figure 2-12, could be analyzed in a similar manner as above to develop 
equations for converting from either the parallel or series mode to the three element 
equivalent circuit that more appropriately represents the physical structure. 



45 

Transmission Electron Microscopy 

The TEM microscope has among the highest lateral spatial resolution in imaging 
mode of any characterization technique. The TEM used for analysis is a Jeol 2010 high 
resolution microscope with a nominal spatial resolution of 1.8A. Additional features of 
the microscope include diffraction information, energy dispersive spectroscopy, and 
electron energy loss spectroscopy. The fundamental basis for the TEM is generation of 
an electron beam from an electron gun. The electron beam then passes through a variety 
of lenses and impinges an ultrathin sample. As seen in Figure 2-13, both scattered and 
unscattered electrons that penetrate the sample thickness comprise the TEM signal. 
Sample requirements for TEM analysis are quite stringent in the sense that they must be 
less than 200nm thick. For the analysis done here, samples were made so that cross 
sectional analysis was capable. Though this type of sample preparation is much more 
complex and delicate than conventional sample preparation, this type of measurement 
gives the ability to determine the thickness of a grown film, which can then be input into 
VASE and XRR models to help develop more realistic and truly physical models. 















46 












2" Stainless 
Steel Heater 



KrF 
Excimer 

Laser 
248 nm 






s 



Collimating Lens 
i * p Aperture 

Focal Lens 



Load Lock 
Doorway 



Load Lock 
Doorway 





Variable Leak 
Valve 



Multi-Target 
Carousel 



Figure 2-1 Ultraviolet-assisted pulsed laser deposition system, KrF excimer laser, and 
optic setup. 



47 




Figure 2-2 Homemade excimer annealing system equipped with vacuum ultraviolet 
lamp. 



48 



Discharge Gap 



HV 

Generator 




Quartz Walls 




uv 

Perforated Outer Electrode 



Gas Inlet 



Figure 2-3 Schematic of excimer lamp illustrating the concentric tube design and how a 
radio frequency load is delivered to the system. Figure adapted from [44]. 



49 




Figure 2-4 Tiny microdischarges from the ignited excimer lamp are the origin of the 
excimer radiation. 






50 



1.00E+08 



1.00E+06 



< 1.00E+04 - 



1.00E+02 



1.00E+00 



Density 



Silicon 

SiOVHfth 

HID: 


Thickness 

GO 

16A 
49 A 


Roughness 

6A 
4A 

4A 


Density 
2.33 g/cm' 
2.73 g/cm-' 
9.46 g/cm 3 



Roughness 



Thickness 



— i 1 — 

2 3 

2-Theta (Degree) 




Figure 2-5 General x-ray reflectivity setup showing physical relationships between the 
acquired data and the modeling output. 












51 












incident 
X-rays 



substrate 



thin film 



diffracted 
X-rays 




Figure 2-6 General x-ray diffraction setup illustrating the interaction of x-rays with a 
structure as they pertain to Braggs Law [56]. 



52 



hv=1486.6eV 




K.E. 



1s 



-*-*- 



■*-*- 



£ 2p ~10eV 
£ 2s ~20eV 






*(eV) 



1s— -*-* 



1s 



-290eV 






Figure 2-7 Process of incoming radiation ejecting a characteristic photoelectron from a 
carbon sample [56] 






53 



image <- 




laser diode 



v / 

PZT scanner 



feedback loop 



Figure 2-8 Schematic of atomic force microscope and the various components that allow 
up to atomic resolution [56]. 






54 



Gate Metal 



Insulator 




Backside Contact 



p-type Semiconductor 



Figure 2-9 Typical MOS capacitor prepared for this dissertation. 









55 



5 



590 
OV Analyzer 

— * 



595 

Quasistatic 

C-V Meter 



H 



5951 

Remote Input 

Coupler 



230-1 
Voltage 
Source 



IEEE-488 BUS 



IT 









U 


Metrics ICS 
Software 


O 


— » 


Computer 
Interface 



Probe 
Station 



Output 



Input 



100k/1M 

Frequency 

Select 




Figure 2-10 Block diagram of Keithley Win-82 system and how it connects to the probe 
station. Adapted from Keithley Win-82 operation manual. 












56 















ox 



Capacitance 




GS 



Onset of Strong Inversion 



V F B V THRESHOLD 

GATE BIAS VOLTAGE, V Q s 



GS 



Figure 2-11 Typical capacitance-voltage illustrating the three main regions that occur in 
a MOS device as a function of bias voltage applied. Adapted from the Keithley Win-82 
operation manual. 



57 



(A) 



JB 



G 



(B) 



jx I — vw 

R 



(C) 



JB 



-AAV 

G 






AAAr 

R 



Figure 2-12 A) parallel, B) series, and C) combined series and parallel models for 
generation of capacitance information. 



58 



Incident beam 




Thin specimen 




Scattered 
electrons 



T 



Unscattered 
electrons 



Figure 2-13 Schematic of electron beam after passing through an ultrathin TEM sample. 
Some electrons are scattered while others remain unscattered [56]. 



59 






CHAPTER 3 
INTERFACIAL LAYER FORMATION 

As mentioned previously, many research teams have been searching for an alternative 
high-k dielectric material to replace the currently used Si0 2 . Although this procedure 
appears to be simple, implementation of this new layer has encountered problems. One 
such problem is the presence of an unwanted interfacial layer that forms between the 
silicon substrate and the alternative high-k dielectric layer. The composition of the 
interfacial layer was and still is a topic of considerable debate with respect to its origin 
and composition. Though it is commonly accepted that oxygen plays an intricate role, 
the exact modes are not well understood. This portion of experimentation looks into 
these questions as they pertain to oxygen conditions in the pulsed laser deposition (PLD) 
system during post deposition heat treatments. 

Anneal Conditions 

For this study, four identical ultrathin Z1O2 samples were grown by conventional PLD 
and then subjected to high temperature anneals in different ambient gases without 
breaking vacuum conditions in the deposition chamber. Due to analysis requirements, a 
special mount was used to attach the silicon by mechanical contact only to the substrate 
heater. This was in contrast to the typical method of attaching the samples with silver 
paste. Because of the alternative mounting technique, there was an increase in thermal 
losses associated with the substrate so even though the substrate heater is capable of 
850°C, annealing temperatures were limited to 750 °C. Of the four samples, the first 
sample was not heat treated in any way so that comparison of an as-deposited sample 



60 

could be conducted. The remaining three samples were annealed for 10 minute 
increments in vacuum, ultra high purity (UHP, 99.999995%) helium, and UHP oxygen. 
Table 3-1 shows the respective partial pressures of oxygen for the various low, mid, and 
high conditions. 



Table 3-1 : Partial pressure o 



Sample 



As-deposited 



Vacuum annealed 



Helium annealed 



Oxygen annealed 



oxygen for respective deposition ambients 



Actual Pressure (Torr) 



5x1 0" 6 



600 



600 



Oxygen Partial Pressure 
(Torr) 



^xTo 7 ^ 



6X10" 4 



600 



Once the samples were processed, they were analyzed by variable angle spectroscopic 
ellipsometry (VASE), x-ray reflectivity (XRR), Fourier transform infrared spectroscopy 
(FTIR), and x-ray photoelectron spectroscopy (XPS). Additionally, a cross sectional 
transmission electron microscopy (XTEM) investigation of the oxygen annealed sample 
was performed. 

A 239A Si02 on silicon calibration wafer was used before any data was collected 
to verify the accuracy of the VASE probe station. When the samples were measured, a 
model with Zr02 on silicon was used. Additional models with an additional interfacial 
layer was considered, however, it was found that more detailed models would return to a 
single layer model when the fitting iterations were allowed to proceed. From the four 
samples, the following data shown in Table 3-2 was collected by VASE. 
There is an increase in overall thickness of the sample as the amount of oxygen in the 
system increases. Since all of the samples were grown under identical conditions, the 



61 



Table 3-2: VASE thickness measurements of Zr0 2 samples after post-deposition heat 
treatments in various ambients. 


Sample 


Thickness (A) 


As-deposited 


39 


Vacuum annealed 


43 


Helium annealed 


47 


Oxygen annealed 


48 



thickness of the Zr0 2 layer is not expected to increase. However, the overall thickness 
may increase if there is an additional second layer that cannot be individually discerned 
by the ellipsometer, but does exist and adds to the overall thickness. 

XRR analysis of the four samples was completed. The data from the VASE 
measurements were used as starting thicknesses for modeling of the XRR spectra. Since 
XRR data takes thickness, roughness, and density of any number of layers into 
consideration, it is of utmost importance to have an approximate idea of the model as it 
pertains to the actual physical sample. As an example of the importance of a good model 
to describe the experimentally acquired data, the oxygen annealed sample is shown in 
Figure 3-1 with a variety of modeling options. 

Figure 3-1 shows the acquired "Raw Data" XRR scan and three different modeling 
possibilities. The first modeling possibility labeled, "1 Layer Model Without Fit" 
represents the situation where the user input values for thickness, roughness, and density 
for a single layer of Z1O2 on top of single crystalline silicon (This would represent the 
ideal case). After inputting the data, no additional processing was conducted. A value 
known as the mean square error (MSE) serves as an indication of the quality of fit 
between the model and the experimental data. The MSE value for the "1 Layer Model 
Without Fit" was -1x10 . As a reference, a poor MSE would be ~lxl0 whereas an 
excellent fit would be around lxlO" 3 . The next modeling possibility labeled, "1 Layer 



62 

Model With Fit" is identical to the previous model, however, additional processing step is 
allowed to occur. The additional processing is an iterative modeling sequence where the 
initial values input from the user are allowed to vary in an attempt to develop a better fit 
(i.e., a lower MSE value). The iterative sequence must be carefully monitored because it 
is possible to obtain a mathematically lower MSE value by losing the physical reality of 
the model. For this example, the density of the Zr0 2 layer has veered from the ideal 
value of 5.68 g/cm 3 and become a physically unrealistic value of 4.77 g/cm 3 . The MSE 
associated with this second model was -1x10"'. The final model labeled, "2 Layer Model 
With Fit" shows that with the addition of a second layer between the Zr0 2 and the silicon 
has a profound positive effect by generating a realistic model exhibiting a very low MSE 
value of 8xlO" 3 . The data for the three different models are shown in Table 3-3. 



Table 3-3: Thickness, roughness, and density data for the various XRR model options 
the oxygen annealed Zr0 2 . 


for 


1 Layer Model 
Without Fit 


Thickness (A) 


Roughness (A) 


Density (g/cm ) 


Silicon Substrate 


00 


5 


2.33 


Zr0 2 Layer 


48 


4 


5.68 










1 Layer Model 
With Fit 








Silicon Substrate 


00 


8 


2.33 




Zr0 2 Layer 


45 


2 


4.77 













2 Layer Model 
With Fit 








Silicon Substrate 


00 


4 


2.33 


Interfacial Layer 


22 


6 


2.40 


Zr0 2 Layer 


16 


4 


5.88 



With respect to the two layer model, initially a density for a pure amorphous Si0 2 
(2.19 g/cm 3 ) [57] layer was input, but this proved inadequate to perfectly describe the 



63 

desired data until the density of the interfacial layer was allowed to vary. Upon variation, 
the density increased to a value of 2.40 g/cm 3 and allowed for the very low MSE value 
shown above. The impact of the higher density value will be discussed later and it will 
be explained why this is still an acceptable physically realistic value. 

Since the importance of the "2 Layer Model With Fit" has been shown, all future XRR 
analysis will implement this type of model. Figure 3-2 shows the four XRR spectra 
corresponding to the as-deposited, vacuum, helium, and oxygen annealed samples and the 
modeling associated with each. The modeled values are shown in Table 3-4. 



Table 3-4: Modeling 
samples. 


data for the as-deposited, vacuum, helium, anc 


1 oxygen annealed 




As-deposited 


Thickness (A) 


Roughness (A) 


Density (g/cm J ) 


Silicon Substrate 


00 


4 


2.33 


Interfacial Layer 


12 


4 


2.68 





Zr0 2 


21 


5 


5.86 










Vacuum Anneal 








Silicon Substrate 


00 


5 


2.33 




Interfacial Layer 


14 


3 


2.67 


Zr0 2 


21 


5 


5.86 










Helium Anneal 








Silicon Substrate 


00 


3 


2.33 


Interfacial Layer 


17 


4 


2.58 


Zr0 2 


20 


5 


5.63 










Oxygen Anneal 








Silicon Substrate 


00 


4 


2.33 


Interfacial Layer 


22 


6 


2.4 


Zr0 2 


16 


4 


5.88 



Figure 3-2 shows that the two layer model was effective in interpreting all of the 
different anneal conditions. From Table 3-4, there are two important features that should 
be noted with respect to the interfacial layer. First, the interfacial layer is the thinnest for 



64 

the as-deposited sample at 12 A and increases to 14, 17, and 22 A when annealed in 
vacuum, helium, and oxygen, respectively. Second, a similar trend is seen for the 
densities of the interfacial layer. In this situation, the density is greatest for the as- 
deposited sample at 2.68 g/cm 3 and decreases to 2.67, 2.58, and 2.40 g/cm 3 when 
annealed in vacuum, helium, and oxygen respectively. Figure 3-3 shows how as the 
partial pressure of oxygen in the system increased, the density of the interfacial layer 
decreased. 

Since XRR investigations indicated that an interfacial layer was present, an XTEM 
sample was prepared from the oxygen annealed sample. This was to determine if the 
micrograph would support the interfacial layer data exhibited in the XRR. As shown in 
Figure 3-4, it is clear that the sample is composed of two distinct layers atop the single 
crystal silicon. Unlike the ideal structure that would exhibit a single Zr0 2 layer atop the 
silicon, there is a distinct polycrystalline Zr0 2 layer atop an amorphous layer atop the 
single crystalline silicon substrate. From the micrograph the thickness of the 
polycrystalline layer is ~20A and the thickness of the amorphous layer is also ~20A. 

The next step in the investigation led to measurement of the samples by FTIR with a 
Nicolet MAGNA 760 instrument in transmission mode. In this case, the backside of the 
samples was cleaned and underwent a 1% hydrofluoric (HF) acid cleansing immediately 
prior to their placement into the nitrogen purged FTIR bench chamber. By using HF on 
the backside of the samples immediately prior to their measurement, this ensured that any 
signal response from the infrared measurement would be not be attributed to Si— O 
bonds on the backside of the oxidized silicon sample and could only originate from the 
dielectric side of the sample (i.e., the interfacial region). Figure 3-5 is a compilation of 



65 

the results recorded from the as-deposited, helium, and oxygen annealed samples. 
Inspection of the plot reveals that the as-deposited control sample already showed a peak 
located around 1080 cm" 1 relating to Si— O— Si bonding in stretching mode. Upon 
annealing in helium, the area under the 1080 cm" 1 peak increased by -93%. When 
annealed in oxygen, the area under the peak increased by -190% with respect to the as- 
deposited peak. Peaks at 800 and 460 cm" 1 corresponding to Si — O — Si bending and 
rocking modes also showed similar increases in absorption. [58, 59] 

The next step in the analysis process included a detailed investigation of the various 
XPS peaks. Figure 3-6 shows the Si 2p peaks as acquired at a take-off angle of 90° 
without any sputtering for the different annealing conditions. There are two prominent 
peaks associated with the Si 2p region. There is one peak that is located at -99.3 eV 
which is attributed to the silicon substrate and a second peak that located at ~103.3eV 
which is attributed to silicon bonded to oxygen in the interfacial layer. While the position 
of the peak located at -99.3 eV remains relatively constant, the position of the higher 
binding energy peak is continually changing as a function of annealing condition. For the 
as-deposited sample the higher energy peak is located at -103.1 eV. Annealing in 
vacuum, helium, and oxygen resulted in a continual increase to values of 103.2, 103.3, 
and 103.5 eV, respectively. The vertical line located at -103.6 eV seen in Figure 3-7 
corresponds to silicon bonded to oxygen in a pure thick SiC>2 layer. [60] 

Analysis of the XPS O Is peaks reveals a single peak exhibiting an asymmetric shape 
indicative of multiple chemical species. In order to analyze this type of peak, a fitting 
program is utilized whereby the single asymmetric peak is deconvoluted into two or more 
Gaussian peaks. An example of the fitting process can be seen in Figure 3-8 which 



66 



shows A) the raw data peak, B) 1 peak fit, C) 2 peak fit, and D) 3 peak fits. Similar to the 
XRR fitting program, the XPS fitting program also has an MSE value that it generates as 
a quality of fit for the different types of models. From Figure 3-8, for the oxygen 
annealed sample, it is clear that a single peak fit is totally unacceptable as well as 
exhibiting an unacceptably large MSE value of 28.5. As seen in B, the two Gaussian 
peak fit has a much better MSE value of 6.5, but is still not considered a good fit. This is 
because the full width at half maximum (FWHM) associated with the peak reaches a 
value of 2.25 eV. Typical values for a peak representing this type of layer will be more 
in the 1 .5 eV range. This leads to the three peak fit shown in C that has a good MSE 
value of 4.8. Additionally, each of the peaks exhibit reasonable values for FWHM and 
can be accounted for in a physical manner. For this example, peak A is corresponded to 
oxygen bound to zirconium, peak B corresponded to oxygen bound to silicon, and peak C 
is associated with trapped physisorbed oxygen present in the pulsed laser deposited films. 
Not all XPS peaks call for three peaks to obtain a good fit. The key it to always have a 
physical model that explains the fit well and also remains within reasonable boundary 
conditions. 

Figure 3-9 shows the three peak fit for each of the four anneal conditions. 
Peaks are located at A) -530.2, B) -531.5, and C) -532.0 eV corresponding to oxygen 
bound to zirconium, silicon, and physisorbed oxygen, respectively. The binding energy 
of the oxygen bonded to zirconium O Is peak matches perfectly with its standard location 
of -530.2 eV in all annealing conditions while the binding energy of the oxygen bonded 
to silicon O Is peak is -1 eV lower than its pure bulk Si0 2 value of -532.5 eV [60]. As 
with the trend seen in the Si 2p peaks, Figure 3-10 shows a similar trend also occurs with 



67 

the binding energies of the oxygen bonded to silicon O Is peak. Again, a continuous 
shift that corresponds directly to the amount of oxygen in the system changes the binding 
energies to values farthest from pure Si0 2 in low oxygen conditions and nearest to pure 
Si0 2 in high oxygen conditions. In addition to the shifting of the peak, there is an 
additional change to note. This second change relates to the size of peak C corresponding 
to trapped oxygen in the film. Analysis of oxygen content reveals that the amount of 
oxygen trapped in the film corresponds to -15% for the as deposited film while all of the 
annealed films reduced to a value of -10%. This important result shows a decrease of 
5% of the trapped oxygen in the film structure when annealed. 

Interpretation of the data reveals several important results from this study. First, 
VASE measurements were conducted as an initial measure of the thickness of the sample. 
While the VASE system was unable to discern discreet differences in the layer, it was 
able to detect an increase of the overall layer thickness. Since the Zr0 2 would not be 
expected to increase during a post deposition anneal, this seemed to indicate growth of a 
second non-Zr0 2 layer. The data from the VASE was used as a first approximation in the 
XRR modeling where the fit revealed that a single Zr0 2 layer atop a silicon substrate 
model was insufficient to describe the shape of the experimentally acquired data. As a 
result, an additional layer was added between the Zr0 2 and the silicon substrate as an 
interfacial layer. The interfacial layer was initially given a density of pure amorphous 
Si0 2 (2.19 g/cm 3 ) [57] which represented the theoretical scenario where the interfacial 
layer is composed entirely of Si0 2 and no other species or compounds are present. 
Without any fitting procedures, this was still an improvement over the single layer model, 
but by allowing the density of the interfacial layer to vary, excellent fits with very low 



68 

MSE values were obtainable for the as-deposited sample and all of the annealed samples. 
The important ramification of the density variations was that the density of the interfacial 
layer was always larger than that of pure amorphous silicon. The values began at a value 
of 2.68 g/cm 3 and decreased toward a more Si0 2 like density as anneal conditions 
incorporated more oxygen. The sample annealed in pure oxygen exhibited an interfacial 
layer of 2.40 g/cm 3 . This important result indicates that the structure of the interfacial 
layer is not a pure Si0 2 , but instead may include higher density materials such as Zr 
metal in a silicate-like structure, or Zr0 2 in a physical mixture with Si0 2 . Other options 
include a layer that consist of a pure Si0 2 , silicate, or silicide, but this cannot be the case, 
since the density of these layers are 2.19, 4.6, and 4.88 g/cm 3 , respectively [57]. As an 
additional confirmation of the existence of the interfacial layer, an XTEM micrograph 
was prepared. Indeed, for the oxygen sample, an amorphous interfacial layer -20 A in 
thickness and a polycrystalline layer of -20 A was observable. This result correlated 
well with the XRR results. The FTIR conducted on the samples revealed a substantial 
increase in area under the 1080 cm" 1 peak (Si — O — Si stretching mode) when the helium 
anneal was conducted and an even larger increase when the oxygen anneal was 
conducted. This increase in absorbance indicates physically more Si — O bonding 
associated with the structure and agrees well with the density and thickness variations 
seen in the XRR data. That is, as the oxygen content in the anneal conditions increase, 
the density of the interfacial layer became more like Si0 2 . XPS investigations of the Si 
2p and O Is regions further supported the existence of the interfacial layer and more 
importantly that it was not composed of a pure Si0 2 . The fact that the Si 2p located 
around -103 eV (instead of 103.6 eV for pure Si0 2 ) shifts to higher binding energies as a 



69 

function of oxygen content indicates that there is either a more complex bonding 
environment for the silicon bonded to oxygen, or a thin film artifact. XRR and FTIR data 
both indicate that a more complex bonding environment more aptly explain this 
discrepancy. Oxygen diffusing to the interface and reacting to form Si0 2 explains the 
change in density seen by the XRR scans and the increase in Si— O bonds exhibited by 
the FTIR spectra. The O Is peak associated with oxygen bonded to silicon is -1 eV away 
from its pure bulk Si0 2 value of -532.5 eV [60]. Similar to the Si 2p peaks, the binding 
energies of the oxygen bonded to silicon continuously shift from values farthest from 
pure Si0 2 at low oxygen conditions to values nearest Si0 2 for the pure oxygen annealed 
sample. 

All of the proceeding data question the source of the oxygen species that make up the 
interfacial layer and the composition of the interfacial layer. The oxygen may come from 
the ambient gases that are used during the heat treatments [61], from trapped oxygen 
located within the grown Zr0 2 film [62-64], or from direct chemical reaction between the 
silicon and the grown layer [65]. 

Due to the predicted thermodynamic stability of Zr0 2 in direct contact with silicon at 
1000 °C [6], a chemical reaction between the grown Zr0 2 film and the silicon substrate is 
unlikely. XPS data supports the migration of trapped oxygen in the Zr0 2 film under high 
temperature treatments. The Ols peak located at -532 eV and corresponding to trapped 
oxygen is seen to decrease from a concentration of -15% in the as-deposited control 
sample to values in the 8 - 10% range for samples that underwent anneals. Growth of the 
interfacial layer during deposition has previously been connected to the amount of 
trapped (physisorbed) oxygen present in the films [64]. However, the amount of oxygen 






70 

that is available in the trapped state is limited. The significant growth of the interfacial 
layer requires an additional oxygen source. The source is oxygen from the ambient 
diffusing through the grown sample, migrating to the interface, and reacting within the 
interfacial region. This is easily an acceptable notion since crystalline Zr0 2 is a relatively 
open structure that is a good oxygen conductor. 

Ultrathin Zr0 2 films were deposited on silicon and then heat treated in different 
atmospheres representing different oxygen conditions. VASE, XRR, XTEM, FTIR, and 
XPS were utilized to determine A) if an interfacial layer was present and B) if there was, 
what role oxygen played in the origin and composition of the layer. It was found that the 
various characterization tools did indeed support the existence of an interfacial, even for 
the as-deposited sample. The composition was determined to be a physical mixture of 
Zr0 2 and Si0 2 that could be affected by oxygen in the system. It was also determined 
that oxygen in the system originated from two sources. The first of these was trapped 
oxygen in the film and the second was from diffusion of ambient gases through the film 
to the interfacial region. Due to these two mechanisms, it has been concluded that it 
would be very difficult to achieve an oxide based alternative high-k dielectric on silicon 
without forming an interfacial layer by conventional PLD. 



71 



A) Raw Data 




C) 1 Layer Model 
With Fit 




.184 0.415 0.646 0.878 1.109 1.340 1.571 1.803 2.034 2.265 2.496 

Theta-2Theta 



D) 2 Layer Model 




Figure 3-1 XRR spectra of A) raw data, B) 1 layer model without fit, C) 1 layer model 
with fit, and D) 2 layer model with fit illustrating the importance of a good model when 
analyzing XRR data. 












72 



A) Oxygen Anneal 




0.179 0.411 0.642 0.874 1.106 1.338 1.569 1.801 2.033 2.265 2.496 

Omega 
Theta-2Theta 



B) Vacuum Anneal 




0.173 0.405 0.638 0.870 1.102 1.335 1.567 1.799 2.032 2.264 2.496 



Theta-2Theta 



Omega 



C) Helium Anneal 




0.195 0.425 0.655 0.885 1.115 1.346 1.576 1.806 2.036 2.266 2.496 

Omega 
Theta-2Theta 



D) Oxygen Anneal 




1E+1 



Theta-2Theta 



Figure 3-2 XRR spectra of A) as-deposited, B) vacuum annealed, C) helium annealed, 
and D) oxygen annealed Zr02 thin films as modeled with the "2 layer model with fit". 



73 



2.8-1 










































2.6 - 




















C 2 4 " 
E 
o 

s 

(A 

C 












Q 2.2 - 
















2.0 - 

1.8 - 















Bulk Si02 



Oxygen 



Helium 



Vacuum 



As-deposited 



Figure 3-3 Plot of interfacial layer density as determined by XRR as a function of 
oxygen content in the annealing system. Bulk SiC>2 has been added as a reference. 



74 




Figure 3-4 Cross sectional TEM of a polycrystalline ZrC>2 thin film atop an amorphous 
interfacial layer atop single crystalline silicon. 


















75 



0.08 



0.07 - 



0.06 - 



0.05 - 



Stretching 



Rocking 




1300 1200 1100 1000 900 



800 



— i 1 1 

700 600 500 400 



Wave Number (cm ) 



Figure 3-5 FTIR spectra of A) oxygen anneal, B) helium anneal, and C) as-deposited 
Zr0 2 thin films showing the increase in Si — O bonding absorption in stretching, bending 
and rocking modes. 



76 






Ideal SiO, 



a 20000 

s 

3 
O 

° 15000 



106 




104 



102 100 

Binding Energy (eV) 



Oxygen 



Helium 



Vacuum 



No Anneal 



96 



Figure 3-6 XPS data of Si 2p region of an as-deposited Zr0 2 thin film and after various 
post deposition anneals. 












77 






103.7 



103.6 



103.5 



> 
• 

* 103.4 

Bf 

E 
■ 

c 
111 

O) 

I 103.3 

c 
1 



103.2 



103.1 



103.0 






Bulk Si02 



Oxygen 



Helium 



Vacuum 



As-deposited 



Figure 3-7 Plot of Si 2p binding energy of oxygen bonded to silicon as a function of 
oxygen content in the annealing system. Bulk Si0 2 has been added as a reference. 



78 











A) Raw Data 


J/-1 










— ' 










c 










3 










o 


- 








U 


jjt 


535 


i I 

5J) 


III » K 




Binding Energy, eV 



Binding Energy, eV 




Binding Energy, eV 



Binding Energy, eV 



Figure 3-8 XPS spectra of A) raw data, B) 1 peak fit, C) 2 peak fit, and D) 3 peak fit fit 
illustrating the importance of a good model when analyzing XPS data. 



79 



c 

o 

U 



A) As-deposited 



Delta = 1.32 eV 




B) Vacuum Anneal 





Binding Energy, eV 






C) Helium Anneal 




/ * \ 


Delta = 1.57 eV 


1 



U 


/ m \\ 








SM 51 




5)7 535 533 531 



d 

3 
o 





Binding Energy, eV 



Delia = 1.22 eV 




Binding Energy, eV 




D) Oxygen Anneal 



Delta - 1.88 eV 



Binding Energy, eV 



Figure 3-9 XPS spectra with "3 peak fit" of A) as-deposited, B) vacuum annealed, C) 
helium annealed, and D) oxygen annealed ZrC>2. 












80 



533.0 



532.5 



532.0 



> 



£ 531.5 

oi 

c 

c 



m 



531.0 



530.5 



530.0 


































' 















1 ! 1 








1 , 1 



Bulk Si02 



Oxygen 



Helium 



Vacuum 



As-deposited 



Figure 3-10 Plot of O Is binding energy of oxygen bonded to silicon as a function of 
oxygen content in the annealing system. Bulk Si0 2 has been added as a reference. 



CHAPTER 4 
ULTRAVIOLET PROCESSING 

Among the results from the experimentation with Zr0 2 post deposition anneals it was 
determined that even the as-deposited sample which did not undergo any additional 
processing already exhibited an interfacial layer. At that stage the interfacial layer was 
determined to have the smallest fraction of detrimental Si0 2 mixed with Zr0 2 compared 
to any later stages of processing, but exhibited the greatest amount of trapped oxygen. 
Since analysis of the interfacial layer revealed that further heat treatments resulted in the 
migration of the trapped oxygen to the interface, this portion of experiments aimed to 
investigate the possibility of reducing the amount of trapped oxygen in an as deposited 
dielectric film through the application of ultraviolet radiations sources. Additionally, this 
experimentation sought to determine if there were any other effects that occurred when 
ultraviolet radiation was utilized in-situ during deposition. A low pressure mercury lamp 
array was added to the conventional PLD system for in-situ ultraviolet-assisted 
deposition. 

In-situ Ultraviolet PLD 

A conventional PLD system was fitted with an array of low pressure mercury lamps. 
Both barium strontium titanate (BST) and yttrium oxide (Y 2 3 ) samples were deposited 
for this study. Barium strontium titanate is an extremely high dielectric constant material 
with reported bulk values ranging from as low as 80 to as high as 3600 in bulk material 
depending on crystallographic orientation [57]. To date, BST has been rejected as a 
possible candidate for MOS applications due to its predicted unstable interface and the 

81 



82 

predicted negative conduction band lineup when in direct contact with silicon. However, 
it still has been studied here due to previous results that indicate positive MOS results, 
and based on its merit as a possible high-k dielectric material in a device that does not 
have such stringent interfacial requirements as a MOS device does. Y 2 3 has also been 
identified as a possible candidate for an alternate dielectric material, but not fully 
embraced because of its moderately high dielectric constant of -10. If pursued, it would 
only provide a solution for a limited number of device generations before encountering 
the same problems Si0 2 is facing now. In this study, the paramount matter was to 
investigate possible effects the ultraviolet radiation may have with respect to properties of 
the deposited film, and specifically how oxygen is present in the system. 

Barium Strontium Titanate (BST) 

Two BST thin film samples were deposited for this study. One sample was deposited 
in the conventional PLD setup while the second was deposited under the UVPLD 
conditions where the low pressure mercury lamp array was utilized. For each case, the 
samples were deposited at a substrate temperature of 650 °C, at a pulse count of 800, and 
a laser fluence of ~1 J/cm 2 . The optimum pressure was determined for each of the two 
samples by growing several samples at different pressures, performing x-ray diffraction 
(XRD) and then comparing the full width at half maximum (FWHM) values of the 
primary (110) peaks. The pressure that resulted in the smallest value for FWHM of the 
major XRD peak was determined to be the most crystalline and used for all subsequent 
experimentation. For BST, the sample deposited without ultraviolet radiation had an 
optimum oxygen deposition pressure of 20 mTorr, while the optimum pressure for the 
sample deposited with ultraviolet radiation had an optimum pressure of only 10 mTorr. 
This turns out to be a very important result that will be discussed in more detail later. 



83 

Under the above mentioned deposition conditions, samples were deposited and 
measured by VASE using a single layer model of barium titanate on silicon. Though the 
deposited samples were actually BaSr .5Tio.50 3 , the model for barium titanate was the 
closest model available and would exhibit similar optical properties to the BST. Any 
attempt to use a two layer model with an additional interfacial layer returned to a single 
layer model when the fitting program was executed. The resulting overall thickness 
values from the single layer model were 475A and 41 5A for UVPLD and PLD deposited 
samples, respectively. These values were used as the first approximations in XRR 
modeling. 

The XRR plots for the PLD and UVPLD deposited samples are shown in Figure 4-1 
and the data for each is shown in Table 4-1. 



Table 4-1 : XRR data for the PLD and UVPLD deposited BST samples. 


PLD Sample 


Thickness (A) 


Roughness (A) 


Density (g/cm ) 


Silicon Substrate 




11 


2.33 


Interfacial Layer 


34 


9 


3.26 


BST 


374 


8 


5.45 


Surface Layer 


18 


6 


3.7 










UVPLD Sample 








Silicon Substrate 




8 


2.33 


Interfacial Layer 


33 


6 


3.45 


BST 


436 


7 


5.46 


Surface Layer 


5 


25 


2.19 



The first important aspect to note in from Table 4-1, is that proper fitting of the BST 
required a "3 Layer Model" to fit the experimentally acquired data. In instances where a 
good fit cannot be readily found with a "2 Layer Model", an additional lower density 
layer may be added to take into account for surface impurities and roughness effects. The 
shape of the XRR plots of the BST, shown in Figure 4-1, revealed compelling evidence 



84 

for the existence of an interfacial layer for both the PLD and UVPLD deposited samples. 
The plot labeled "UVPLD Raw Data" shows a dampened region from -0.425 to -0.675° 
along the theta-2theta axis where the amplitude of the oscillations did not match all of the 
other values along the theta-2theta axis. The only method to account for this type of 
dampening is with the addition of an interfacial layer exhibiting a particular thickness and 
density combination. The plot labeled "UVPLD 1 Layer Model" shows how the 
oscillations remain constant along the entire theta-2theta axis without the addition of the 
interfacial layer. Sections C and D of Figure 4-1 show the best fits that could be achieved 
with the "3 Layer Model" for the UVPLD and PLD deposited samples, respectively. In 
the case of the PLD deposited sample, the thickness of the interfacial layer was -34 A 
whereas the UVPLD sample was 33 A. More importantly, the density of the PLD 
deposited sample was -3.26 g/cm 3 whereas the UVPLD was -3.45 g/cm 3 . This value is 
greater than that of a pure amorphous Si0 2 , but still much less than a pure silicate or 
silicide. Another important characteristic to note included the high surface roughness of 
25 A associated with the UVPLD sample versus the only 6 A for the PLD sample. 
Finally, the overall thicknesses from XRR are in good agreement with the earlier values 
determined from VASE. 

In order to verify the existence of the interfacial layer seen in the XRR results, an 
XTEM micrograph of a BST film was obtained and is shown in Figure 4-2. The features 
of the micrograph include a two layered structure composed of a polycrystalline BST 
layer atop an amorphous interfacial layer atop the single crystalline silicon. Though the 
polycrystallinity of the BST layer is not apparent in Figure 4-2, less magnified images do 
in fact reveal the polycrystalline morphology. 



85 

In an effort to verify the large roughness differences found between the PLD and 
UVPLD deposited samples, atomic force microscopy (AFM) measurements were made 
on the samples. Seen in Figure 4-3, it was found that the AFM measurements made over 
an area of 1/rni x lftm verified the XRR results showing RMS values of 7 A to 23 A for 
PLD and UVPLD deposited samples, respectively. The AFM measurements also 
revealed a great deal of surface texturing associated with the UVPLD deposited sample 
with an average grain size of -300 nm. This is in contrast to the PLD sample which 
exhibits neither surface texturing nor large grains. 

The data for the XRD investigation of both the UVPLD and PLD deposited samples 
are shown in Table 4-2. 



Table 4-2: XRD peak 


data for the PLD and UVPLD deposited BST samples. 


PLD Sample 


2-Theta Peak 
Location (°) 


FWHM (°) 


Peak Orientation 


Peak 1 


31.70 


0.5573 


(110) 


Peak 2 


45.45 


0.5568 


(111) 










UVPLD Sample 








Peak 1 


31.81 


0.4705 


(110) 


Peak 2 


45.61 


0.5691 


(111) 



The XRD spectra were composed of two distinct peaks for each. The main difference 
between the PLD and UVPLD deposited samples was the slightly smaller (-0.0816° less) 
FWHM value in the (110) peak for the UVPLD deposited sample. An additional 
difference between the two samples is a slight shift in the 2-theta values between the PLD 
and UVPLD samples. These differences were attributed to both the grain size increase 
seen in the UVPLD sample as well as stresses inherent in the thin film system. A 
variation of XRD, grazing incidence x-ray diffraction (GIXD), was also completed for 



86 



the two samples. The GIXD plots are shown in Figure 4-4 and the data is shown in Table 



4-3. 



Table 4-3: GIXD peak data for the PLD and UVPLD deposited BS 



PLD Sample 



Peak 1 



Peak 2 



Peak 3 



Peak 4 



Peak 5 



Peak 6 



UVPLD Sample 



Peak 1 



Peak 2 



Peak 3 



[ samples: 



2-Theta Peak 
Location (°) 



31.74 



39.10 



45.68 



51.18 



56.55 



66.28 



31.84 



FWHM (°) 



0.5292 



0.5481 



0.6273 



0.5850 



0.6997 



0.7341 



0.5172 



39.27 



Peak 4 



Peak 5 



Peak 6 



45.53 



51.30 



56.67 



660.47 



0.5281 



0.6692 



Peak Orientation 



(110) 



(111) 



(200) 



(210) 



(211) 



(220) 



(110) 



(HI) 



0.7095 



0.6256 



0.6930 



(200) 



(210) 



(211) 



(220) 



Due to the different manner in which a GIXD plot is acquired, there are 6 distinct 
peaks associated with the scan. Matching the scans to the JCPDS diffraction file the scan 
indicates a good quality cubic Bao.s Sro. 5 Ti0 3 . with all listed peaks in the JCPDS being 
accounted for in the spectrum. Inspection of the 2-theta positions of the peaks indicates 
slight shifting of the peaks to higher values for four out of the six orientations when 
ultraviolet radiation was used during processing. This shift was also accompanied by 
lower FWHM values for all of the associated orientations. Two of the six orientations, 
the (200) and (210), shifted to lower 2-theta values and were accompanied by higher 
values of FWHM. 

The XPS data was taken without any sputtering and at a take-off angle of 90°. The 
plots of the O Is region revealed a need for three peaks to properly fit the experimentally 



87 

acquired data. The binding energies associated with the peaks are A) -530.1, B) -532.0, 
and C) -533.5 eV for oxygen bound to barium, oxygen physically trapped, and other 
surface impurities, respectively. A detailed inspection of peak B shows the key 
difference between the PLD and the UVPLD technique. As seen in Figure 4-5, with the 
application of the UV radiation, the amount of trapped oxygen is reduced from a value of 
-15% to a value of only -10%. This is of particular importance since the application of 
ultraviolet radiation during deposition results in less trapped oxygen available in the 
material that is available to migrate to the interface and react. 

Two additional samples were deposited under identical conditions with the exception 
of a lower pulse count of 500. The resulting thickness as determined by VASE and XRR 
were 250 A and 232 A for UVPLD and PLD deposited samples, respectively. This pair 
of samples was made into MOS devices so differences in the electrical properties could 
be determined. To properly determine a good device for testing, current- voltage 
measurements were made on the two samples. Figure 4-6 shows a plot of current density 
versus voltage and that both of the samples exhibited very low leakages of 1 x 1 A/cm 
over a wide range of voltages from minus 3 to positive 1 .5 volts. The high frequency 
capacitance versus voltage plot is shown in Figure 4-7. It shows how the accumulation 
capacitance associated with the UVPLD deposited sample is enhanced compared to that 
of the PLD deposited sample. With a dot size of 8.08x1 0" 8 m 2 the capacitances 
associated with the PLD and UVPLD deposited samples are 662 pF and 790 pF, 
respectively. Using equation 1-7, the equivalent oxide thickness of the conventionally 
grown PLD sample is reduced from an actual thickness of 232 A to an EOT of 42 A. 
More importantly, the UVPLD sample reduced from an actual thickness of 250 A to and 



88 

EOT of 35 A! These relatively low EOT values are a result of overall dielectric constant 
calculations of 21.5 and 27 for the PLD and UVPLD deposited samples, respectively. 
This represents a 25.5% increase in dielectric constant with the application of ultraviolet 
radiation. 

For the in-situ UVPLD study of BST, VASE measurements were completed for a base 
measurement of thicknesses that were input as a starting value for XRR modeling. The 
XRR revealed two important findings. The first was the existence of an interfacial layer 
with a density greater than that of pure Si0 2 and the second was a large surface roughness 
associated with the UVPLD deposited sample, but not the PLD deposited sample. An 
XTEM investigation verified that indeed there was an amorphous interfacial layer while 
AFM verified that the UVPLD deposited sample exhibited a great deal of texturing with 
large -300 A grains. The increased texturing was also indicated in the XRD and GLXD 
through more intense peaks with lower FWHM values for certain peaks while others 
became less intense with larger FWHM values. XPS was able to discern a reduction in 
the amount of physically trapped oxygen associated with the UVPLD deposited sample. 
Reasoning for this lower amount of trapped oxygen stems from the lower oxygen 
pressure that is enabled during a UVPLD deposition as a result of the addition of the 
ultraviolet radiation sources. This lower amount of oxygen can also explain the greater 
density seen in the interfacial layer for the UVPLD deposited sample since the ratio of 
Si0 2 to BST would be less. Finally, capacitance voltage measurements revealed an 
enhanced dielectric constant and low EOT values for the UVPLD deposited sample. 
While the low EOT values were not less than the current 1 8 A Si02 layer used, it did 
clearly illustrate two important notions. First, it shows how a high dielectric constant can 



89 

take a thick (relatively speaking) layer not subject to tunneling leakage issues and allow it 
to perform like an ultrathin layer. Secondly, it illustrates the important role that 
ultraviolet radiation can play in achieving high quality dielectrics when deposited under 
more reactive oxygen conditions. 

All of the above mentioned results question the claim that BST is a poor selection as 
an alternative dielectric material because of its thermodynamic instability with silicon 
and its lack of a conduction band barrier. The answers to these two questions are as 
follows: The calculations for band offsets generated by Robertson [5] were made for a 
given dielectric when in an ideal direct contact with silicon. However, it was mentioned 
that if there were some type of buffer layer between a dielectric and the silicon that this 
would change the band offset properties. This is what had happened with the BST. As 
seen in the XTEM image of BST, there is clearly an interfacial layer that had formed. As 
an extension of the results from the Zr0 2 study, it is reasonable to conclude that the BST 
had also generated an interfacial layer that was composed of a physical mixture of SiC>2 
and BST. With this extension, the band offset associated with the structure would be 
intermediate between the value of BST in direct contact with silicon and that of S1O2 in 
direct contact with silicon thus eliminating that issue. Similarly, the lack of 
thermodynamic stability of BST in direct contact with silicon is also questionable since 
the interface is in fact an interfacial layer that is partially Si0 2 . It is the unusually high 
dielectric constant of BST that still allowed the multilayered system to generate such 
good dielectric constant values and low equivalent oxide thicknesses. It also follows that 
if the interfacial layer had even less SiC>2 or was physically thinner, the effect of the low- 
k layer would be less pronounced allowing even better results. With respect to the 



90 

structural changes seen in the data, energetic photon bombardment of the surface would 
most rationally explain these results. More importantly, the fact that the ultraviolet 
radiation is acting in an in-situ manner allows the photons to continually interact with 
each atomic layer as the structure is deposited. This added energy could account for the 
structural differences noted. 

For the BST, ultraviolet radiation has been used to modify chemical properties of the 
thin film system by reducing the amount of trapped oxygen in the deposited layer. Less 
trapped oxygen in the structure during deposition and cooldown dictates less possible 
migration of oxygen species to the interface during deposition and post deposition 
processes resulting in less negative effects with respect to the interfacial layer. The effect 
of energetic photons that impinge each layer of the material during deposition can 
account for the structural changes that were noted in the crystallinity and roughness. The 
combined effect of the chemical and structural changes resulted in enhanced electrical 
properties with good leakage currents and low EOT values. It is concluded that due to a 
more controlled oxygen environment, as a result of application of ultraviolet radiation, 
higher quality samples can be deposited which exhibit superior structural, chemical and 
electrical properties. 

Yttrium Oxide (Y 2 3 ) 

Ultraviolet radiation studies were also conducted for yttrium oxide. While it has 
been passed over by many research teams as a possible alternative dielectric layer, most 
of the rationale for this decision is due to the theoretical dielectric constant of -10 that 
Y2O3 exhibits. In this case, replacement of Si02 would only be effective for a few 
generations before Y2O3 faced many of the same issues that Si02 faces today. 
Regardless, it has been studied here in an attempt to determine the effects of ultraviolet 



91 

radiation on an additional oxide material that satisfies the criteria for a good candidate. 
To start the investigation, samples were grown in a similar manner as the BST whereby 
optimum growth pressures were determined for a given growth temperature. In this case, 
the Y 2 3 thin films were grown for a pulse count of 300 at a temperature of 400 °C and 
an oxygen pressure of 4 mTorr and 10 mTorr for UVPLD and PLD, respectively. Once 
again, an array of low pressure mercury lamps was used in-situ during the deposition for 
the UVPLD deposited sample. 

VASE investigations of the UVPLD and PLD deposited samples measured overall 
thicknesses of -85 A and -76 A, respectively. Much like the Zr0 2 and BST samples, a 
single layer model was the most appropriate for modeling the data. This is once again 
attributed to the inability of VASE to discern additional layers without returning to a 
single layer approximation, but not necessarily an indication that there is not actually a 
second layer present. 

The XRR plots are shown in Figure 4-8 and the data from the model is shown in Table 
4-4. 



Table 4-4: XRR data for the PLD and UVPLD deposited Y 2 3 


sam 


pies. 


PLD Sample 


Thickness (A) 


Roughness (A) 


Density (g/cm ) 


Silicon Substrate 


00 


5 


2.33 


Interfacial Layer 


16 


6 


2.27 


Y2O3 


50 


3 


5.03 


Surface Layer 


5 


4 


3.61 










UVPLD Sample 








Silicon Substrate 


00 


5 


2.33 


Interfacial Layer 


15 


4 


2.52 


Y2O3 


56 


5 


5.03 


Surface Layer 


7 


4 


3.60 






92 

A 3 layer model was used to fit the experimental data. This created in similar results to 
the BST study with the need for an interfacial layer in the modeling to appropriately 
describe the acquired data. The interfacial layer for the UVPLD deposited sample 
exhibited a density of -2.52 g/cm 3 whereas the density of the PLD deposited sample was 
a lower value of -2.27 g/cm 3 . Both of these interfacial layers had a density greater than 
that of pure amorphous Si0 2 and less than that of any type of yttrium silicate or silicide. 

An XTEM micrograph was obtained for the Y 2 3 system to determine if an interfacial 
layer was present. As shown in Figure 4-9, the Y 2 3 exhibited a polycrystalline Y 2 3 
layer atop an amorphous interfacial layer atop the single crystalline silicon. This result is 
comparable to the Zr0 2 and BST results previously shown. The XTEM image also 
showed a columnar structure associated with the polycrystalline Y 2 3 film. This was the 
case for both PLD and UVPLD grown samples. 

An investigation of the crystallinity of the two samples by GIXD, seen in Figure 4-10, 
revealed a single peak associated with Y 2 3 that is described in Table 4-5. 



Table 4-5: GIXD 


peak data for the PLD and UVPLD deposited Y 2 3 


samples 


PLD Sample 


2-Theta Peak 
Location (°) 


FWHM (°) 


Peak Orientation 


Peak 1 


30.34 


1.3052 


(222) 










UVPLD Sample 








Peakl 


30.21 


1.8045 


(222) 



GIXD was chosen over XRD for this set of experiments since the thickness of the layer 
was quite low and GIXD would be a more effective mode to observe meaningful 
differences in the resulting peaks. A single peak that corresponded to a (222) oriented 
cubic Y 2 3 layer was present. The FWHM value for the UVPLD deposited sample was 
smaller than the PLD deposited by -0.493° which is once again an indication of larger 



93 

gain sizes. Additionally, the position of the (222) peak shifted -0.13° indicating stresses 
associated with the thin film system. Inspection of Figure 4-9 will reveal that there are 
additional peaks located around -50 -55°, but these peaks have all been attributed to 
silicon peaks such as the (311). 

Inspection of the XPS data also reveals a similar result as the BST. In the case of the 
Y 2 3 , the O Is peaks, shown in Figure 4-11, the asymmetric peak can be deconvoluted 
into two distinct peaks. Peak A is located at -529.1 eV and peak B is located at -530.9 
eV. Peak A located at 529.1 eV corresponds to yttrium bound to oxygen while peak B 
located at -530.9 eV corresponds to trapped oxygen in the layer. Similar to the BST, 
there is a clear reduction in the amount of trapped oxygen with the application of 
ultraviolet radiation. The value reduces from -20% trapped in the conventional PLD 
grown sample to -10% in the UVPLD sample. 

It is thus shown once again how ultraviolet radiation played a direct role in altering the 
chemical properties of the oxide system reducing the amount of trapped oxygen within 
the Y2O3 structure. Once again, similar to BST, XRR investigations revealed the need 
for an interfacial layer with a greater density than pure amorphous Si0 2 to properly fit the 
XRR spectra. The interfacial layers existence was verified by XTEM while GIXD 
indicated marginal differences in grain structure and/or internal film stresses. XPS 
investigations showed how ultraviolet radiation during deposition allowed for lower 
ambient oxygen pressures which resulted in a marked reduction of trapped oxygen in the 
films. 



94 



A) UVPLD Raw Data 




0.233 0.459 0.686 0.912 1.138 1.365 

Theta-2Theta 



1.591 1.817 2.044 2.270 2.496 



B) UVPLD 1 Layer 
Model 




0.233 0.459 0.686 0.912 1.138 1.365 

Theta-2Theta 



1.591 1.817 2.044 2.270 2.496 



C) UVPLD 3 Layer 
Model 




0.233 0.459 0.686 0.912 1.138 1.365 

Theta-2Theta 



1.591 1.817 2.044 2.270 2.496 



D) PLD 3 Layer Model 




0.259 0.483 0.706 0.930 1.154 1.378 

Theta-2Theta 



1.601 1.825 2.049 2.273 2.496 



Figure 4-1 XRR spectra of A) UVPLD raw data, B) UVPLD 1 layer model, C) UVPLD 
3 layer model, and D) PLD 3 layer model for BST samples. 



95 




Figure 4-2 Cross sectional TEM of a polycrystalline BST thin film atop an amorphous 
interfacial layer atop a single crystal silicon substrate. 



96 



PLD 













w 




UVPLD 






Figure 4-3 AFM of BST for PLD and UVPLD deposited samples. The UVPLD 
deposited sample exhibits increased roughness larger grain sizes. 









97 









ra 

c 

< 



c 

ZJ 
O 

O 



25 



(110) 




(HI) 




•*+<*****-*•* * .>** 



30 35 40 









(200) 





(210) 



(211) UVPLD 

(220) 



L. j\ A ^ A 

^*W«/VfW*-i*T A**-^*********^ S^*^«w W « # Jr#* r ^Oy^^^^fc-^ v 



PLD 



^K t> ■*■ »<■■. «.^.*^.^*H — s* | 



-i 1 - 



45 50 

2-Theta (Degrees) 



55 60 65 70 



Figure 4-4 GIXD of BST for PLD and UVPLD deposited samples. The UVPLD 
deposited sample exhibits increased (110) texturing. 



98 




Binding Energy (eV) 



lOt 



CO 



o 

U 




Binding Energy (eV) 



Figure 4-5 O Is XPS of BST for PLD and UVPLD deposited samples. Peak B 
corresponds to the amount of physically trapped oxygen in the thin film structure and is 
reduced in the UVPLD deposited sample. 



99 



100E+00 



1 OOE-01 




1 00E-06 



1 

Voltage (V) 



Figure 4-6 Current density versus voltage plot for BST of PLD and UVPLD deposited 
samples. 









100 



900E-10 -j 






UVPLD 


8 00E-10 - 








7 00E-10 - 






/^^ PLD 


6.00E-10 - 




j 




E 

• 500E-10 ■ 

1 4 00E-10 ■ 
o 




Y 






I 




3 00E-10 ■ 








2 00E-10 - 








1 OOE-10 - 
OOOEtOO - 




1 

1 1 1 


— i 1 1 



-3 



1 

Voltage (V) 



Figure 4-7 High frequency capacitance versus voltage plot for BST of PLD and UVPLD 
deposited samples. 



101 



A) UVPLD Raw Data 




0.126 0.363 0.600 0.837 1.074 1.311 

Theta-2Theta 



1.548 1.785 2.022 2.259 2.496 
Omega 



B) UVPLD 3 Layer 
Model 




0.126 0.363 0.600 0.837 1.074 1.311 

Theta-2Theta 



1.548 1.785 2.022 2.259 2.496 
Omega 



C) PLD Raw Data 



c 

o 
U 




0.126 0.363 0.600 0.837 1.074 1.311 

Theta-2Theta 



1.548 1.785 2.022 2.259 2.496 
Omega 



D) PLD 3 Layer 
Model 




0.126 0.363 0.600 



0.837 1.074 1.311 
Theta-2Theta 



1.548 1.785 2.022 2.259 2.496 
Omega 



Figure 4-8 XPS spectra of A) UVPLD raw data, B) UVPLD 3 layer model, C) PLD raw 
data, and D) PLD 3 layer model for Y 2 3 samples. 






102 










Figure 4-9 Cross sectional TEM of a polycrystalline Y 2 3 thin film atop an amorphous 
interfacial layer atop a single crystal silicon substrate. 



103 



400-r 



(222) 



50: 



f^HUi^MW* 



^^^ WvA^^^^ 




UVPLD 



O 1 ' ' ' ' i ' ' ' ' I ' ' ' ' I ' ' ■ ■ I ' ' ' ' I ' ' ' ' I ' ' ' ' I ' 



1 1 1 1 1 1 1 1 1 ■ 1 1 ■ ■ 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 ■■ ■ i ■ i ■■ i ■■ i ■ i ' 

45 50 55 60 65 



30 



35 40 



3 2Theta 



Figure 4-10 GIXD of Y 2 3 for PLD and UVPLD deposited samples. The UVPLD 
sample exhibits increased (222) texturing. 



104 




529 

Binding Energy (eV) 



Figure 4-11 Ols XPS of Y 2 3 for PLD UVPLD deposited samples. Peak B corresponds 
to the amount of physically trapped oxygen within the thin film structure. 



CHAPTER 5 
POST DEPOSITION ULTRAVIOLET ANNEALING 

The final area of research related to the post deposition annealing of Hf0 2 thin films. 
In this experimentation, the goal was to determine if there were any ultraviolet effects 
that resulted from adding an ultraviolet radiation source to post deposition oxygen anneal. 
Since the films were amorphous and particularly thin, there were several possibilities that 
could occur. The first was that the oxygen would directly react with the H1O2 amorphous 
network to better oxygenate the film. The next was the possibility of the high energy 
photons being emitted from the lamp changing the structural characteristics (amorphous 
to crystalline) of the films. Finally, there was the possibility of oxygen passing directly 
through the films and reacting with the silicon at the interface and growing an Si0 2 
interfacial layer. 

Ultraviolet- Assisted Oxidation of Silicon 

Since the ultraviolet radiation source for this portion of the experimentation was the 
home built silent barrier discharge excimer lamp, a preliminary test to verify the effect of 
ultraviolet radiation and its ability to generate atomic oxygen and ozone was completed. 
In this study, a series of samples had their native Si0 2 layer removed by a dip in 1 % 
hydrofluoric acid for 10 minutes. They were then immediately loaded into a high 
vacuum system equipped with the excimer ultraviolet radiation source. Two sets of four 
samples were processed whereby the samples were oxidized for 30, 60, 90, and 120 
minutes in 5 mTorr of oxygen at 300 °C with and without ultraviolet radiation present. 
At this temperature dry oxygen annealing was not expected to generate a Si0 2 layer of 

105 



106 

any appreciable thickness. In contrast, previous work by Boyd et. al. [29] indicated that 
if intense ultraviolet radiation was being emitted from the excimer source, dry oxygen 
would be converted into atomic oxygen and ozone which would lead to appreciable 
oxidation of the silicon. The results are displayed in Table 5-1 and shown graphically in 
Figure 5-1. 



Table 5-1 VASE thickness measurements for silicon oxidized with and without 
ultraviolet radiation at 300°C. 


Time (minutes) 


Thickness (A) 


UV Thickness (A) 


30 


5 


35 


60 


6 


46 


90 


7 


56 


120 


7 


63 



VASE measurements verified that indeed, there was not appreciable oxidation of the 
silicon with dry oxygen annealing where the greatest thickness of Si0 2 was only 7A for 
the longest duration anneal. As a comparison, the sample annealed in the presence of 
ultraviolet radiation for the same time duration was oxidized to 63A. This nine-fold 
increase in thickness verified that the excimer source was working properly. 

Hf0 2 Post Deposition Anneal 
Since the oxidizing effectiveness of the ultraviolet source had been proven, H1O2 
samples were annealed at various temperatures with and without ultraviolet radiation. 
The goal was to determine the role that UV radiation could play in a post deposition 
anneal versus using a dry oxygen anneal. The samples were grown by Motorola 
Corporation by a chemical vapor deposition setup and delivered directly for analysis (i.e., 
the samples did not undergo any of the "typical" densification anneals). The samples 
were amorphous and grown to a thickness of ~55A of which, ~10A was attributed to an 
interfacial layer. This data was supplied by Motorola without additional elaboration 



107 

about their deposition or measurement techniques. Seven samples were processed for 
this experimentation. One sample was set as an as-deposited control sample, while the 
other six samples were annealed at 300, 475, and 600°C in 5 mTorr of oxygen with and 
without ultraviolet radiation. XRD, GIXD, AFM, VASE, XRR, and capacitance-voltage 
measurements were conducted on these samples. 

XRD of the as-deposited control sample showed that the structure of the films was 
amorphous. That is, there were not any distinguishable peaks typically associated with a 
crystalline Hf0 2 structure. More importantly, none of the samples that underwent post 
deposition heat treatments showed any diffraction peaks. This indicates that all of the 
Hf0 2 films regardless of whether ultraviolet radiation was present retained their 
amorphous character. GIXD was also conducted as a secondary check to this finding 
since the films were quite thin. GIXD also revealed retention of the amorphous state 
throughout all anneal conditions with and without ultraviolet radiation. 

AFM measurements were conducted and, unlike in-situ UV irradiation, there were no 
changes found, such as texturing, in the surface morphology. All samples exhibited a 
random morphology with RMS roughness values in the 5 - 7 A range. This was 
consistent with the lack of crystal linity noted in the samples from the XRD and GIXD 
data. All of these findings indicate that there were not any structural changes associated 
with the Hf0 2 system. 

VASE measurements were made for each of the anneal conditions and are shown in 
Table 5-2. From Figure 5-2, it can be seen that dry oxygen annealing did not result in an 
increase of the overall thickness of the HfC>2 layer for any of the anneal temperatures. 



108 



Table 5-2 VASE thickness measurements for Hf0 2 anneals with and without ultraviolet 
r^Hiatinn , 




Temperature (°C) 


Thickness (A) 


Excimer Ultraviolet 
Radiation 


Sample 1 


As-deposited 


52 


— 


Sample 2 


300 


53 


NO 


Sample 3 


475 


52 


NO 


Sample 4 


600 


53 


NO 


Sample 5 


300 


56 


YES 


Sample 6 


475 


59 


YES 


Sample 7 


600 


64 


YES 



On the other hand, the anneals conducted with excimer ultraviolet radiation continually 
increased in thickness from a starting value of -53 A for the as-deposited sample to a 
value of -64 A for the 600°C anneal. These values were used as a first approximation in 
the XRR modeling. 

XRR measurements of the samples indicated slight differences in the thickness of the 
total film thickness that compared well with the values seen in the VASE measurements. 
A "3 Layer Model" was used for the interpretation since an interfacial layer was required 
for good model fits to be found. The thickness of the Hf0 2 layer remained relatively 
constant, but there were distinct variations in the thickness and density of the interfacial 
layer. These variations are presented in Table 5-3 and shown in Figures 5-3 and 5-4. 

Figure 5-3 clearly shows that the samples annealed in dry oxygen had only a marginal 
increase in the thickness of the interfacial layer, but only for the sample that was annealed 
at the highest temperature. However, those samples annealed in the presence of 
ultraviolet radiation had an increase in the interfacial layer thickness even for the 300 °C 
anneal. The samples that were annealed at higher temperatures had more substantial 
increases in the thickness of the interfacial layer. Figure 5-4 shows that the density of the 
interfacial layer is at its highest value for the as-deposited sample and that any subsequent 



109 



Table 5-3 XRR data for the thickness and density of the amorphous interfacial layer- 



Sample 1 
Sample 2 



Sample 3 



Sample 4 



Sample 5 



Sample 6 



Sample 7 



Temperature 
(°C) 



As-deposited 



300 



475 



600 



300 



475 



600 



Excimer 

Ultraviolet 

Radiation 



NO 



NO 



NO 



YES 



YES 



YES 



Interfacial 
Layer 
Thickness (A) 



15 



15 



15 



16 



18 



20 



23 



Interfacial 
Layer 
Density (g/cm 3 ) 



2.84 



2.79 



2.67 



2.42 



2.71 



2.50 



2.34 



heat treatments corresponded to decreases in the density. The samples that were annealed 
in the presence of ultraviolet radiation decreased in density faster than the dry oxygen 
anneal samples for all of the anneal temperatures. For example, the density of the sample 
annealed at 475 °C without ultraviolet radiation had a density of 2.67 g/cm 3 while the 
sample annealed with ultraviolet radiation at the same temperature had decreased to 2.5 
g/cm 3 . 

Capacitance voltage measurements were made on all samples. As mentioned 
previously, the accumulation capacitance will be a function of the Hf0 2 and any 
interfacial layers that are present in the system. If there is a change in the oxygen 
environment in the system, this will be reflected in the accumulation capacitance of the 
whole system. Capacitance- voltage results for the system are shown in Figure 5-5. It can 
be seen that the as deposited sample exhibited the greatest value of accumulation 
capacitance. All subsequent anneals resulted in a decrease in the accumulation 
capacitance. The accumulation capacitance of the samples that were annealed in the 
presence of ultraviolet radiation decreased more quickly than the dry oxygen annealed 
samples for any given temperature. 



110 

In summary, structural analysis was completed in the form of XRD, GIXD and AFM. 
For each of these analyses, there were not any measurable differences between the as- 
deposited sample and any of the annealed samples. XRD and GIXD verified retention of 
the amorphous state of the Hf0 2 and AFM showed no changes in the morphology of the 
surface. VASE was used to determine an overall thickness that was used as the starting 
points for XRR investigations. XRR verified the existence of an interfacial layer and 
showed that as the annealing temperatures increased, variations in the density and 
thickness of the interfacial layer occurred. The important ramification of this finding was 
that the samples annealed in the presence of ultraviolet radiation decreased in density 
more rapidly than the dry oxygen annealed samples while the thickness of the interfacial 
layer increased more rapidly than the dry oxygen annealed samples. Capacitance-voltage 
measurements exhibited the highest accumulation capacitance for the as deposited sample 
and all subsequent anneals resulted in lower accumulation capacitances as a function of 
temperature and ultraviolet illumination. 

The explanation for these results relates to the ability of the ultraviolet radiation to 
convert dry oxygen to atomic oxygen and ozone. The atomic oxygen can migrate more 
easily within the HfD 2 structure. This ability to easily migrate accompanied by 
increasing temperature allows the oxygen to migrate to the silicon interface, react, and 
form Si0 2 . In the as-deposited sample, the density of the interfacial layer is greatest 
since there has not been any additional oxygen added to the system. That is, the mixture 
of Si0 2 and Hf0 2 that exists has the smallest fraction of Si0 2 at this stage. Any 
additional anneals increases the Si0 2 fraction resulting in decreasing the density to values 
that are closer to pure Si0 2 as well as contributing to the growth of the interfacial layer 



Ill 

overall. This growth of the interfacial layer manifests itself readily in the electrical 
characterization where the low-k dielectric layer plays an increasingly important role. 

The main conclusions from this study are that post deposition annealing in an oxygen 
ambient will have a detrimental effect on the electrical properties of the device. The 
application of ultraviolet radiation amplifies the problem by creating more mobil highly 
reactive oxygen species that can move easily to the interface and react with silicon. If 
there are any beneficial oxygenation effects that occur within the Hf0 2 layer, their benefit 
is quickly nullified by the detrimental effects of the low-k interfacial layer that develops 
during the heat treatments. 












112 
















20 



40 



60 80 

Time (min) 



100 



120 



Figure 5-1 VASE thickness versus time results for oxidation of silicon with and without 
excimer radiation in oxygen. 



113 




100 



200 



300 400 

Temperautre (°C) 



500 



700 



Figure 5-2 VASE thickness of the overall Hf0 2 for samples annealed with and without 
ultraviolet radiation. 



114 



25 



20 



s 

E 



15 - 



10 




100 






200 



300 400 

Temperature (*C) 



500 



600 



700 



Figure 5-3 XRR thickness measurements of the Hf0 2 interfacial layer for samples 
annealed with and without ultraviolet radiation. 






115 



Black = UV 
Gray = no UV 




Temperature (°C) 



Figure 5-4 XRR density measurements of the Hf0 2 interfacial layer for samples 
annealed with and without ultraviolet radiation. 






116 



4.50E-10 

4 00E-10 

3 50E-10 ■ 

3 00E-10 

g 2.50E-10 ^ 

c 

I 

g. 2 00E-10 
is 

o 

1 50E-10 

100E-10 

5 00E-11 - 



OOE+00 




-25 



-25 



-2 



UVPLD 

A) As-received 

B) 300 °C 

C) 475 °C 

D) 600 °C 



-1.5 -1 



-0 5 5 

Voltage (V) 



1.5 



-1.5 



-0.5 05 

Voltage (V) 



1.5 



25 



4.50b- 10 - 




PLD 


4 00E-10 - 


A)___^ 


A) As-received 


3 50E-10 - 


C) ^--^ 


\ B) 300 °C 




^\ C) 475 °C 


3 00E-10 - 


n** 


-^\\ D) 600 °C 




L*) 


St 

• 2 50E-10- 

c 






S 2 00E-10 - 




1 


o 




1 


1 50E-10 - 






1 00E-10 - 




\\ 


5.00E-11 - 






0.00E+00 - 


1 1 T 1 -T 1 



2.5 



Figure 5-5 Capacitance-voltage measurements of HfD 2 MOS devices for samples 
annealed with and without ultraviolet radiation. 









CHAPTER 6 
CONCLUSIONS 

Since Si0 2 will be unable to continue as the dielectric material of choice in MOSFET 

device structures, there exists a driving force to replace the existing Si0 2 layer with an 

alternative layer with a higher dielectric constant. Though many research teams have 

investigated numerous different materials in an attempt to accomplish this goal, to date, 

none have succeeded due to new problems associated with the insertion of the alternative 

layer. The more prominent of these issues include the formation of an unwanted low-k 

interfacial layer that forms between the silicon substrate and the dielectric material as 

well as excessive leakage currents. The studies associated with this dissertation 

addressed the issues related to these two problems as they are related to oxygenation in 

the growth system. 

First, experimentation was conducted whereby ultrathin Zr0 2 samples were deposited 
and exposed to low, medium, and high oxygen annealing conditions at 750 °C. From the 
various characterization techniques, it was determined that the films deposited by pulsed 
laser deposition exhibited an interfacial layer even for samples that were not post- 
deposition annealed. Additionally, post deposition annealing under high vacuum 
conditions resulted in development of the interfacial layer and reduction of physically 
trapped oxygen in the film structure. FTIR analysis of post deposition annealed samples 
in helium and oxygen resulted in substantial increases in the Si — O bonding. The final 
conclusion from this portion of experimentation was that the interfacial layer is already 
formed during deposition. Subsequent annealing results in a migration of physically 

117 



118 

trapped oxygen in the film to the interface that reacted wilh silicon and increased the 
Si0 2 content in the interfacial region. Since this amount of oxygen was limited, it was 
also concluded that oxygen in the ambient deposition gases also migrated to the interface 
to react with the interface. 

In the second portion of experimentation, ultraviolet radiation from a low pressure Hg 
lamp was added to the conventional pulsed laser deposition system. The addition of 
ultraviolet radiation to the growth system was selected because of its ability to react with 
dry oxygen to create atomic oxygen and ozone. These species are more reactive than dry 
oxygen and, in the case of the atomic oxygen, can move easily through the lattice of a 
dielectric film. It was found from the optimization process for deposition of the BST and 
Y 2 3 that UV radiation allowed for lower oxygen pressures to be used during deposition 
while still attaining high quality samples. The impact of this finding is of significant 
relevance because it results in less physically trapped oxygen in the deposited structures 
and less oxygen available for reaction during deposition. Additionally, it was found that 
several other properties such as texturing and electrical characteristics were altered with 
the application of UV radiation. It was found that there still existed an interfacial layer 
from deposition alone, but the characteristics of that interfacial layer were altered as a 
results of the different oxidizing species generated by the ultraviolet radiation. This 
played an important role in the growth of good quality films. 

In the final portion of experimentation, already deposited H1D2 thin films received 
from Motorola Corporation were exposed to excimer ultraviolet radiation. In doing so, It 
was found that the changes in structure that were seen for the in-situ irradiation did not 
occur for the post deposition annealed samples. Capacitance voltage measurements 



119 

indicated that there were more reactive species present in the ultraviolet annealed samples 
compared to the samples processed without ultraviolet radiation by the fact that the 
overall accumulation capacitances of the CV plots continually decreased. In this 
scenario, the oxygen species migrate immediately through the Hf0 2 layer to the interface 
and add to the growth of the interfacial layer. 

Studies in this dissertation indicated that there is development of an interfacial layer 
during deposition and that ultraviolet radiation could be beneficial to the growth system. 
The addition of UV radiation enhanced the properties of the thin film if used for in-situ 
deposition; however, it can also be detrimental with respect to devices if migrating 
species are allowed to migrate either from the film or from the ambient to the interface 
and react with silicon to further develop the low-k interfacial layer. 






APPENDIX 
ELECTRICAL DATA EXTRACTION METHOD 

When conducting electrical characterization, once the proper model has been chosen 
for the with the Metrics ICS software, measurements may be acquired and a significant 
amount of information may be extracted from this data. For example, by calculating the 
total fixed charge in a MOS stack, numerous other important data are generated from the 
process. The derivation of the fixed charge will be done as an example. 

It is important to first get a current-voltage measurement from one of the MOS devices 
to determine an appropriate diode to make capacitance-voltage measurements on. That 
is, a diode with a high breakdown voltage and turn on characteristics. Once this is 
determined, a high frequency/quasistatic output from the Win-82 system may be 
generated. A typical result for a p-type wafer is shown in Figure 2-11. While both 
curves are important, for this process, the calculations will be based off the high 
frequency curve. The first step is to use the CV curve to calculate doping concentration 
in the substrate. This is done by plotting the reciprocal of the square of the capacitance 
versus the gate voltage. The new plot should exhibit a linear region. By taking the slope 
of the linear region, it is possible to calculate the doping in the semiconductor. It is 
important to note that curvature in this region may represent nonuniform doping in the 
semiconductor. The equation for determining the doping is shown below: 

N = i- (A-l) 

qs s A (slope) 



120 



121 



where q is the electric charge, s s is the semiconductor dielectric constant, A is the gate 
area, and TV is the doping (acceptor or donor). With the value, it is now possible to 
calculate the Debye length (L D ). 



j £skT (a-2) 

where k is the Boltzmann constant and T is temperature in Kelvin. With this value the 
semiconductor flatband capacitance can be calculated. 

C„ s =^ <A-3) 

The next step is to calculate or experimentally determine the capacitance of the dielectric 
layer (Q). In the scenario described here, the capacitance has been measured and 
represents the value of capacitance of the high frequency curve in the accumulation 
region. This typically is taken at a value of minimal leakage current, as ascertained by 
the previous IV measurements. If, one was unable to measure the capacitance, but had 
knowledge such as the thickness of the dielectric layer, d, area of the MOS stack, A, and 
the dielectric constant, s t , of the dielectric layer, then it is possible to calculate the 
corresponding capacitance of the insulator. The relationship is shown as follows: 

C ( . = ^ (A-4) 

The next important parameter in the process is calculation of the overall flatband 
capacitance: 

C C 

w /roc*-', 



c -FBS^i (A . 5) 

c +c 



and the ratio 



122 
^FB (A-6) 

Ci 

The next step is to calculate the V shift this is done by plotting C/Q versus gate voltage. 
Then use the ratio of C F B/Cjto to visually determine the flatband voltage shift (AV sh m). In 
an ideal MOS capacitor, the flatband voltage condition corresponds to a AV sh ift = 
thereby allowing the conclusion that any voltage shift should be related to total fixed 
charge. This charge may come from many different sources: 

Q m > Mobile Charges > location dependent on bias 

Q ot > Oxide Trapped > distributed in the oxide 

Q f > Fixed Oxide > near surface charge 

Q it > Interfacial Traps > at the interface 

The first three of these charges causes shifts of the overall CV curve to the left or right 
while the fourth charge manifests itself as changes in slope (stretching) of the overall CV 

curve. 

The next step is to determine the value <£ms, which is the work function of the metal 
minus the work function of the semiconductor. Depending on the wafer type used, this 
would be done as follows 

E. 



p-type <t> MS =<Pm-<P =<P M - 



n-type tp MS =</> M -(/> =<t>, 



2 



E 

2 B 



(A-7) 



(A-8) 



where E g /2 is the intrinsic Fermi level, $m is the metal work function, </>s is the 
semiconductor work function, X is the electron affinity of the semiconductor, and Tb is 
the Fermi level difference between the Fermi level and the intrinsic Fermi level. 






123 



Finally, the total oxide charge may be calculated with the following equation by 
collating all of the previously determined values as: 



n-type N f = 



AV-</> 



MS 



(A-9) 



<7 

Another important electrical value that is commonly cited in literature is the value of the 
interface traps, Qit. There are several possibilities for determining this value: 1) 
comparison of the high frequency capacitance to a theoretical capacitance without any 
interface traps, 2) comparison of a low frequency capacitance to a theoretical capacitance 
without any interface traps, or 3) comparison of a high frequency capacitance to a low 
frequency capacitance. The third option, first developed by Castagne and Vapaille 
consisted of combining the low and high frequency curves (i.e., the output of the Win-82 
system) to determine the silicon surface capacitance per unit area, C s . Unlike the Terman 
method where only the high frequency capacitance (C H f) is measured, or the Berglund 
method where only low frequency capacitance (C L f) is measured, this method measures 
both C LF and Chf- This eliminates the need for the generation of any of the theoretical 
computations needed for the other methods. As outlined in more detail by Nicollan and 
Brews [65], the basic result is that interface trap level can be determined at a given 
voltage as a function of position in the bandgap by simple input of measured parameters 
into the following equation: 



ft-4£ 



ii 



\ C HF +AC 

Cox ) 



r 

1 _ HF 



^ox ) 



(A-10) 



where Dit is the interface trap density, C ox is the capacitance per unit area, and 

AC = C HF -C LF (A-ll) 

This equation is used to make a plot if D it0 versus Chf/C ox where 






A, - A, 



124 



1000 (A . 12) 



and x is the thickness of the oxide in question. Combination of the directly measured 
CV data and the above equation it is therefore possible to determine the Dito. It is 
important to realize, however, that accurate values of Dit cannot be determined over the 
entire range of gate bias' due to a variety of reasons including round-off errors, errors due 
to use of a 1 Mhz CV curve, and error from C LF . Each of these errors is addressed in 
detail in Nicollian and Brews. 












LIST OF REFERENCES 

1 . S. M. Sze, Physics of Semiconductor Devices, 2 nd ed. (Wiley, New York, 1 98 1 ) 

2. D. A. Buchanan: IBM J. Res. Develop. 43, 245 (1999) 

3. E. H. Nicollian, and J. R. Brews, MOS (Metal Oxide Semiconductor) Physics and 
Technology (Wiley, New York, 1982) 

4. G. D. Wilk, R. M. Wallace, J. M. Anthony: J. Appl. Phys. 89 (10), 5243 (2001) 

5. J. Robertson: J. Vac. Sci. Technol. B 18 (3), 1785 (2000) 

6. K. J. Hubbard, D. G. Schlom: J. Mater. Res. 11(11), 2757 (1996) 

7. M. Copel, M. Gribelyuk, E. Gusev: Appl. Phys. Lett. 76 (4), 436 (2000) 

8. M. Houssa, M. Naili, C. Zhao, H. Bender, M. M. Heyns, A. Stesmans: Semicond. Sci. 
Technol. 16,31-38(2001) 

9. M. E. Hunter, M. J. Reed, N. A. El-Masry, J. C. Roberts, S. M. Bedair: Appl. Phys. 
Lett. 76(14), 1935(2000) 

10. S-K. Kang, D-H. Ko, E-H. Kim, M. H. Cho, C. N. Whang: Thin Solid Films 353, 8 
(1999) 

1 1. S. J. Wang, C. K. Ong, S. Y. Xu, P. Chen, W. C. Tjiu, A. C. H. Huan, W. J. Yoo, J. S. 
Lim, W. Feng, W. K. Choi: Semicond. Sci. Technol. 16, LI 3 (2001) 

12. J. Gottmann, A. Husmann, T. Klotzbucher, E. W. Kreutz: Surf. Coat technol. 100- 
101,415(1998) 

13. Jun-Ying Zhang, I. W. Boyd: Appl. Phys. Lett 71 (20), 2964 (1997) 

14. A. G. Grigoryants, Basics of Laser Material Processing (Mir Publishers, Moscow, 
1994) 

15. A. El-Habachi, K. H. Schoenbach: Appl. Phys. Lett. 73 (7), 885 (1998) 

16. A. El-Habachi, K. H. Schoenbach: Appl. Phys. Lett. 72 (1), 22 (1998) 



125 



126 



17. D. Lindau, H. F. Dobele: Rev. Sci. Instrum, 59 (4), 565 (April 1988) 

18. M. Yoshimoto, K, Takubo, M. Komoda, H. Matsunami: Appl. Surf. Sci 79/80, 264- 
269 (1994) 

19. Y. I. Nissim, J. M. Moison, F. Houzay, C. Licoppe, M. Bensoussan: SPIE 1393, 216 
(1990) 

20. V. Craciun, A. Srivastave, J. M. Howard, R. K. Singh, J. Perriere: Appl. Phys. A 69 
[Suppl.],S787(1999) 

21. A. Srivastava, V. Craciun, J. M. Howard, R. K. Singh: Appl. Phys. Lett. 75 (19), 3002 
(1999) 

22. A. Kazor, I. W. Boyd: Appl. Phys. Lett. 63 (18), 2517 (1993) 

23. P. Bergonzo, P. Patel, I. W. Boyd, U. Kogelschatz: Appl. Surf Sci. 54, 424-429 
(1992) 

24. P. Bergonzo, I. W. Boyd: Appl. Phys. Lett. 63 (13), 1757 (1993) 

25. J-Y. Zhang, I. W. Boyd: Electr. Lett. 32 (22), 2097 (1996) 

26. J-Y. Zhang, I. W. Boyd: Appl. Phys. Lett. 77 (22), 3574 (2000) 

27. J-Y. Zhang, B. Lim, 1. W. Boyd, V. Dusastre: Amer. Inst. Phys. 73 (16), 2299 (1998) 

28. J-Y. Zhang, I. W. Boyd: J. Appl. Phys. 80 (2), 633 (July 1996) 
29. 1. W. Boyd: Mater. Sci. Forum 173/174, 81-92 (1995) 

30. S. Kubodera, M. Kitahara, J. Kawanaka, W. Sasaki, K. Kurosawa: Appl. Phys. Lett. 
69 (4), 452 (1996) 

31. R. S. Taylor, K. E. Leopold, K. O. Tan: Appl. Phys. Lett. 59 (5), 525 1991 

32. C. Manfredotti, F. Fizzotti, M. Boere, G. Piatti: Appl. Surf. Sci. 69, 127 (1993) 

33. J. Flicstein, Y. Vitel, O. Dulac, C. Debauche, Y. I. Nissim, C. Licoppe: Appl. Surf. 
Sci. 86,286-293(1995) 

34. O. B. Postel, M. A. Cappelli: Appl. Phys. Lett. 76 (5), 544 (2000) 

35. A. Yokotani, N. Takezoe, K. Kurosawa: Appl. Phys. Lett. 69 (10), 1399 (1996) 



127 

36. J. D. Ampetepe, J. Diggs, D. M. Manos, M. J. Kelley: J. Appl. Phys. 85 (11), 7505 
(1999) 

37. J-Y. Zhang, S. L. King, I. W. Boyd, Q. Fang: Appl. Surf. Sci. 109/110, 487-492 
(1997) 

38. J-Y. Zhang, H. Esrom, I. W. Boyd: Appl. Surf. Sci. 96-98, 399-404 (1996) 

39. J-Y. Zhang, I. W. Boyd, H. Esrom:Surf. & Interface Anal. 24, 718-722 (1996) 

40. J-Y. Zhang, I. W. Boyd: Electr. Lett. 33 (10), 911 (1997) 

41. A. Kazor, I. W. Boyd: Appl. Surf. Sci. 54, 460-464 (1992) 

42. H. Imai, K. Awazu, M. Yasumori, H. Onuki, H. Hirashima: J. Sol-Gel Sci. & 
Technol. 8, 365 (1997) 

43. J-Y. Zhang, L-J. Bie, I. W. Boyd: Symp. B, E-MRS Spring Meeting (1997) 

44. U. Kogelschatz: Appl. Surf Sci. 54, 410 (1992) 

45. G. Boehme, Atlantic Ultraviolet Corporation Brochure (Hauppauge, New York, 
1996) 

46. B. Eliasson, P. Erni, M. Hirth, U. Kogelschatz: United States Patent 4,837,484 
(1989) 

47. M. Aguilar-Frutis, M. Garcia, C. Falcony: Appl. Phys. Lett. 72 (14), 1700 (1998) 

48. B. H. Lee, Y. Jeon, K. Zawadzki, W-J Qi, J. Lee: Appl. Phys. Lett. 74 (21), 3143 
(1999) 

49. V. Craciun, R. K. Singh: Appl. Phys. Lett. 76 (14), 1932 (2000) 

50. B. K Moon, C. Isobe, J. Aoyama: J. Appl. Phys. 85 (3), 1731 (1999) 

51. A. Yokozawa, A. Oshiyama, Y. Miyamoto, S. Kumashiro: IEEE IEDM 97-703 
(1997) 

52. S. Maruno, T. Kuroiwa, N. Mikami, K. Sato, S. Ohmura, M. Kaida, T. Yasue, T. 
Koshikawa: Appl. Phys. Lett. 73 (7), 954 (1998) 

53. Y. Matsui, K. Torii, M. Hirayama, Y. Fujisaki, S. Iijima, Y. Ohji: IEEE Electron Dev. 
Lett. 17 (9), 431 (1996) 






128 

54. D. Kumar, J. Sankur, K. G. Cho, V. Craciun, R. K. Singh: Appl. Phys. Lett. 77 (16), 
2518(2000) 

55. H. Wengenmair, J. W. Gerlach, U. Preckwinkel, B. Stritzker, B. Rauschenbach: Appl. 
Surf. Sci. 99, 313 (1996) 

56. C.R. Brundle, C. A. Evans, S. Wilson: Encyclopedia of Materials Characterization 
(Butterworth-Heinemann, Boston, 1992) 

57. CRC Handbook of Chemistry and Physics, 78 th ed., edited by D. R. Lide (CRC Press, 
Boca Raton, 1997) 

58. D. A. Neumayer and E. Cartier, J. Appl. Phys. 90, 1801 (2001) 

59. W. S. Lau, Infrared Characterization for Microelectronics (World Scientific, 
Singapore, 1999) 

60. Handbook of X-ray Photoelectron Spectroscopy, edited by J. F. Moulder, W. F. 
Stickle, P. E. Sobol, and K. D. Bomben Physical Electronics, Inc. (Eden Prairie, 
Minnesota, 1986) 

61. H. Ono, Y. Hosokawa, T. Ikarashi, K. Shinoda, N. Ikarashi, K. Koyanagi, and H. 
Yamaguchi, J. Appl. Phys. 89, 995 (2001) 

62. t. j. Vink, R. G. F. A. Verbeek, J. H. M. Snijders, and Y. Tamminga, J. Appl Phys. 87, 
995 (2001) 

63. V. Craciun, J. M. Howard, N. D. Bassim, and R. K. Singh, Mater. Res. Soc. Symp. 
Proc. 666 (2002) 

64. V. Craciun, J. M. Howard, N. D. Bassim, and R. K. Singh, Appl. Surf. Sci. 168, 123 
(2000) 



BIOGRAPHICAL SKETCH 
Joshua Michael Howard was born July 8 th , 1975, in Park Ridge, Illinois. He grew up 
in Crystal Lake, Illinois, a small northwest suburb of Chicago. Upon graduation from 
Crystal Lake Central High School in 1993, he began an undergraduate curriculum at the 
University of Missouri-Rolla in Ceramic Engineering. In 1997, after completing a 
bachelor of science in Ceramic Engineering and a minor in Spanish, Joshua moved to 
Gainesville, Florida, to begin his graduate career at the University of Florida in the 
Department of Materials Science and Engineering. He joined Dr. Rajiv K. Singh's 
research team in the spring of 1998, and began a project to develop a novel high intensity 
excimer based vacuum ultraviolet radiation source and then apply the technology to pulse 
laser deposition of high dielectric constant thin films on silicon. He received his master's 
degree in 1999 and continued on in pursuit of a PhD. Post graduation plans include 
working as a post doctoral researcher with Dr. Singh on additional excimer based studies. 



129 



I certify that I have read this study and that in my opinion it conforms to 
acceptable standards of scholarly presentation and is fully adequate, in scope and quality, 
as a dissertation for the degree of Doctor of Philosophy. 

r 



U 



v ;; l { - 



Rajiv K. Singh, Chairman 
Professor of Materials Science and 
Engineering 



I certify that I have read this study and that in my opinion it conforms to 
acceptable standards of scholarly presentation and is fully adequate, in scope and quality, 
as a dissertation for the degree of Doctor of Philosophy. , 



A Aj (au 



Stephen J. JPearton — "^ 
Professor of Materials Science and 
Engineering 



1 certify that I have read this study and that in my opinion it conforms to 
acceptable standards of scholarly presentation and is fully adequate, in scope and quality, 
as a dissertation for the degree of Doctor of Philosophy. 



C // , : , . 



i-h\ 



Cammy Abertiathy 
Professor of Materials Science and 
Engineering 



1 certify that I have read this study and that in my opinion it conforms to 
acceptable standards of scholarly presentation and is fully adequate, in scope and quality, 
as a dissertation for the degree of Doctor of Philosophy. 

(J 




David P. Norton 

Associate Professor of Materials Science 
and Engineering 



















tf^l 



UNIVERSITY OF FLORIDA 



III 



3 1262 08555 3435